Interview guides · Bechtel

Bechtel Software Engineer Interview Guide

Bechtel Software Engineer interview guide: three reported rounds, resume and integration design questions, behavioral STAR prep and a 7-day plan.

The 60-second brief

Bechtel is an engineering, procurement, construction and project management company, usually shortened to EPC. Candidates describe its engineering software teams as building and maintaining software systems, data pipelines, automation tools and enterprise tools for infrastructure, energy, nuclear, defense and mining projects. Examples that come up in candidate reports include digital twin applications, integrations with process modeling tools, workflow platforms built on ServiceNow, and modernisation of legacy enterprise databases. A Software Engineer in this setting works next to discipline leads such as civil and mechanical engineers, project managers and field superintendents, turning their operational needs into specifications, web applications, APIs and data integrations.

This guide is for candidates applying as a Software Engineer at Bechtel, from university and entry-level hires to experienced engineers. It covers the three reported rounds (Initial Screening, Technical Interviews, Behavioral Questions) and the 19 questions candidates report, which fall into four groups: resume and past-project deep dives, software design and systems integration, behavioral and situational judgment, and practical case scenarios such as a slow field tool or a spreadsheet migration.

Candidates report three rounds over roughly three to five weeks: an initial screening that includes a phone interview with HR, technical interviews run by engineering managers or team leads, and a behavioral round that mixes behavioral and technical questions. This reflects candidate accounts, not an official schedule.

Focus areas

Process Modeling
Engineering Design (Detailed Engineering)
P&ID Diagrams (Piping & Instrumentation Diagrams)

Overview

The work behind this role is software for engineering and construction projects: web applications, backend APIs, data integrations between enterprise platforms and discipline-specific tools, and upkeep of legacy databases. That shapes the reported questions. Candidates describe a loop that leans on resume discussion, architecture scenarios, tool integration and practical problem-solving more than on isolated algorithm puzzles. Prepare accordingly: for every project on your CV, be able to draw the architecture, name the data flows, and defend each technology choice against an alternative you rejected.

The stack that candidates see listed is broad. Languages named are C#, Java, Python and JavaScript or TypeScript, with relational database design, SQL optimisation, REST services (and SOAP for older integrations) and common design patterns as the core. Cloud platforms (AWS or Azure), Docker, CI/CD, React or Angular, ServiceNow and MATLAB scripting appear as extras. You do not need all of them. Pick the ones on your resume, refresh the details you could be asked to explain, and say plainly when a tool is new to you and how you would learn it.

The 19 reported questions split into resume deep dives (an architecture walk-through, language and platform choices, third-party integration, a production debugging story, schema design and query tuning), design questions (a real-time site-data pipeline, monolith versus microservices, design patterns, consistency with intermittent connectivity, a multi-discipline reporting dashboard), behavioral questions (conflicting stakeholders, deadlines and scope changes, a mistake, disagreeing with a senior colleague, an unfamiliar domain or location) and case scenarios (replacing a spreadsheet workflow, diagnosing latency, speed versus test coverage, access control for sensitive documents). The design and case questions reward the same habits: state assumptions, list failure modes, and say what you would measure.

Context matters in this loop. The users are field engineers and discipline leads, connectivity at remote sites can be poor, and reports mention that defense, government and nuclear teams may raise security clearance and that mobility or relocation can come up early. Tie your answers to those facts: offline behaviour, data integrity, audit trails, and explaining trade-offs to people outside software.

The Bechtel interview process

Candidates describe three stages, from an HR screen to technical interviews to a behavioral round. Use the descriptions below to decide what to rehearse for each.

Initial Screening

Candidates report a phone interview with HR to check basic qualifications. Some also describe an automated video screen such as HireVue instead of, or alongside, the call, and for early-career applicants in certain regions a written assessment on engineering and logical fundamentals followed by a group discussion. Topics reported for this stage are your background, mobility and location preferences, and salary alignment. For defense, government or nuclear teams, candidates are advised to be ready to discuss security clearance eligibility. Approach it as a short, prepared conversation: concise career story, clear logistics, no surprises.

What to demonstrate

Basic qualifications against the role, as the stage is described
Your professional background as it appears on the resume
Mobility, travel and location preferences, and expectations on pay

How to prepare

Write a short career summary that links your two or three strongest projects to enterprise software, integrations or data work.
Decide your position on relocation, site travel and temporary assignments, and state it plainly instead of leaving it open.
Settle a pay range from your own research before the call, and know your clearance eligibility if you apply to government, defense or nuclear teams.
If the screen is a recorded video, rehearse a two-part answer (what you did, what resulted) out loud into a camera without notes.

Technical Interviews

Candidates report technical interviews run by engineering managers or team leads, sometimes as one or two panels held on Microsoft Teams or at a regional office, with principal engineers, project engineering leads and hiring managers mentioned. The reported emphasis is resume discussion, system architecture scenarios, tool integration and practical problem-solving for engineering projects, rather than algorithm-only exercises. Expect to cover the question types in this guide: past-architecture and technology-choice questions, third-party integration such as ServiceNow or ERP systems, schema design and query tuning, data pipelines, monolith versus microservices, offline consistency, and case scenarios such as a slow production tool or access control for sensitive documents.

What to demonstrate

Technical skills, as the round is described, including the choices behind your past projects
Software design and integration reasoning for engineering and data workflows
Structured problem-solving on diagnostic and migration scenarios

How to prepare

For each major resume project, prepare a one-page architecture sketch with the data flow, the datastore, and one decision you would change now.
Outline an integration with a third-party platform (ServiceNow or an ERP) covering auth, retries, idempotency, error handling and how you test it.
Practise a schema and slow-query walkthrough: read an execution plan, add or reshape an index, and explain what stays slow at ten times the rows.
Rehearse one latency investigation and one access-control design aloud, naming the first measurement or the first role you would define.

Behavioral Questions

Candidates describe this round as a mix of behavioral and technical questions that probes interpersonal skills. Reported behavioral themes are working with non-technical stakeholders, handling deadlines and scope changes, owning a mistake, disagreeing with a senior colleague, and adapting to an unfamiliar domain or location. Because technical questions are mixed in, treat each story as something you may be asked to open up technically: which system, which failure, which fix. The STAR structure (situation, task, action, result) is the format candidates are advised to use.

What to demonstrate

Interpersonal skills, as the round is described
How you work with colleagues from non-software backgrounds
Ownership and learning in stories about mistakes and disagreements

How to prepare

Build six stories, one for each reported behavioral prompt, each with a decision you personally made and a result you can state with numbers where you have them.
For every story, keep one technical detail ready (the system, the root cause, the fix) in case the conversation turns to specifics.
Practise explaining a technical trade-off to a civil, mechanical or process engineer without jargon, using schedule, cost and safety effects.
Write two questions for the panel about their stack and how the software team works with field operations.

Where candidates lose points

Walking the resume as a list of technologies with no reasons behind them

For each project, write down the problem, the architecture, why you chose that language, framework and datastore over one alternative, and one thing you would change. Reports describe detailed resume questions, so every line on your CV needs this.

Designing the integration or site-data pipeline for the happy path only

When you outline an integration with ServiceNow, an ERP or a site-data pipeline, add the cases that break it: duplicate delivery, malformed or oversized files, a vendor API timeout, and a field site that is offline for hours. Name idempotency keys, validation at the edge, a dead-letter store and how a client resyncs.

Answering a latency or crash scenario with a list of guesses

Start with what changed and when, then measure: where the time goes (client, API, database, network), whether load or data volume explains it, and how to reproduce. State the first metric you would pull and what each result would rule out before suggesting a fix.

Telling behavioral stories in the team's voice, with no decision or result of your own

Write each story in STAR form with the sentence that starts with 'I decided' and a result you can quantify. Cover all reported themes: conflicting stakeholders, deadlines, mistakes, disagreement and unfamiliar domains, with a different story for each.

Staying vague on mobility, relocation or clearance at the HR screen

Reports say mobility, location and clearance eligibility can come up in the first stage and affect onboarding. Decide your position before the call and state it clearly, along with the pay range you researched.

Questions & practice

Archive a resource graph without breaking live references or recursing

Medium · graph traversal · topological order · tenant isolation

Resources reference other resources within a tenant; for the largest tenant the reference table holds up to 2,000,000 nodes and 8,000,000 edges. Archiving a resource must archive everything reachable from it that nothing outside the set still references, refuse when a live external referrer exists, and terminate when references form cycles, which they legitimately do. Produce the archive order and the refusal list, targeting O(V+E). Say what stops the traversal crossing a tenant boundary, and why recursion is the wrong control structure at this size.

Approach

Load the subgraph with the tenant predicate on both endpoints of the edge, not only on the side you started from. Scoping the left table alone is the classic cross-tenant leak: one mis-entered edge then pulls another tenant's resources into the traversal and, worse, into the archive.
Traverse iteratively with an explicit stack. A 2,000,000-node graph can hold a chain deep enough to exhaust a native stack in the low tens of thousands of frames, and that failure is a process crash rather than an error you can return.
Treat cycles as data rather than corruption: compute strongly connected components with Tarjan in O(V+E) using its own explicit stack, then condense. The condensation is a DAG, so a topological order over it gives the archive order, and every member of a component archives in one transaction because no order within a cycle is valid.
Decide refusals with reverse edges. A candidate is archivable only if every in-edge originates inside the candidate set, so build the transpose or count in-degrees restricted to the visited set, and emit each blocked resource with the id of the external referrer, which is the only part of the answer an operator can act on.

Follow-up

The graph is read in one query and the archive writes a minute later. What can change in between, and how do you make the write safe?
The candidate set is 400,000 resources. Is that one transaction, and if not, what does a half-finished archive look like to a reader?

Canonicalise a request body into a stable idempotency fingerprint

Medium · parsing · canonicalisation · hashing

idempotency_key.request_fingerprint is a SHA-256 over the method, path and canonicalised body, and a retry whose fingerprint differs must be rejected with 422 rather than served the stored response. Write the canonicaliser. Bodies are JSON up to 256 KB nested at most 32 levels; clients vary key order, whitespace and unicode escaping, and some send 64-bit ids as JSON numbers. Produce a deterministic byte string such that semantically identical bodies match and any semantic difference does not. State your complexity and name two normalisations you refuse to perform.

Approach

Parse once into a tree, then re-serialise under fixed rules: object keys sorted, array order preserved, one escaping convention, no insignificant whitespace. Parsing is O(n) and sorting keys is O(k log k) per object, so O(n log n) overall with O(depth) stack, and the 32-level cap is enforced during parsing because hostile nesting is how a canonicaliser becomes a stack overflow.
Sort keys by their UTF-8 bytes and say why the obvious implementation is wrong in some runtimes: a default string comparison that orders by UTF-16 code units places surrogate pairs, meaning code points from U+10000 up, below U+E000 to U+FFFF, which is not UTF-8 byte order, so two services written in different languages disagree on the same document.
Do not re-encode numbers through a double. IEEE-754 binary64 represents integers exactly only up to 2^53, so normalising a 19-digit id through a float changes it, and 1 against 1.0 cannot be reconciled without deciding whether they are the same value. Preserve the literal token, and require ids as strings at the API boundary if you want them comparable.
Reject duplicate keys rather than picking one. JSON permits them and parsers disagree, most keeping the last, so any choice you make ties the fingerprint to a parser detail that the code handling the request does not necessarily share.

Follow-up

A client sends the same logical request with an extra field your API ignores. Same key, different fingerprint, so you return 422. Is that the right answer?
Where does the fingerprint get computed relative to request decompression and the body-size limit?

Find overlapping job attempts and peak concurrency from lease records

Medium · sweep line · intervals · leases

A day of job_run history yields about 50,000,000 attempt records: (job_run_id, job_type, attempt, started_at, finished_at which is NULL when the worker died, lease_expires_at). Leases expire on a clock, so a job that outran its lease ran twice. Produce (a) every job_run_id whose attempts overlapped in wall-clock time and (b) the peak number of simultaneously running attempts per job_type with the minute it occurred. Target O(n log n). State how you treat a NULL finished_at and what clock skew does to your answer.

Approach

Define the interval before sorting anything: an attempt occupies [started_at, COALESCE(finished_at, lease_expires_at)). finished_at is observed and lease_expires_at is only a promise, so every attempt without a finish contributes an estimate and the whole result is a lower bound on overlap rather than an exact count.
For peak concurrency, sweep: emit 2n endpoints, sort by (timestamp, kind) with ends ordered before starts at equal timestamps, then walk the sequence maintaining a counter per job_type and record each type's maximum with its timestamp. O(n log n) dominated by the sort, O(n) space, or O(1) extra if the sort is external and the walk streams.
For overlap detection, do not compare attempts pairwise. A single global sort by (job_run_id, started_at) gives both the grouping and the order; within a group, keep the maximum end seen so far and report an overlap exactly when the next start is less than that running maximum, which is one linear pass after the sort.
Half-open intervals matter and are easy to get wrong: with closed intervals an attempt ending at the same millisecond another begins reads as concurrency two, and across 50,000,000 records that artefact swamps the real signal.

Follow-up

A handler is not idempotent and you have found 400 overlapping jobs. Which of them actually caused damage, and what would you query to find out?
Peak concurrency for one job_type is 4 against a configured cap of 4. Is the cap working, or is the data hiding attempts that never started?

How do you approach database schema design and query optimization when handling high-volume operational data?

Medium · Technical Experience & Resume Deep Dive

Approach

Start from the access patterns: list the top read queries and the write paths. Normalise the transactional tables (around 3NF) and add denormalised or summary tables only for read hotspots you have measured.
Get keys, types and constraints right: surrogate primary keys, foreign keys with an index on the referencing column (neither PostgreSQL nor SQL Server creates that index automatically), DECIMAL for quantities and money, timezone-aware timestamps, and NOT NULL and CHECK constraints so the database protects data integrity.
Index for the predicates: in a composite index put equality columns before range columns, use covering indexes (INCLUDE in PostgreSQL 11+ and SQL Server), and keep indexed columns out of functions. Rewrite WHERE YEAR(created_at) = 2024 as a half-open date range so it stays sargable.
Diagnose from the plan rather than by guessing: EXPLAIN (ANALYZE, BUFFERS) or the actual execution plan. Look for full scans of large tables, estimated vs actual row counts far apart (refresh statistics), sorts spilling to disk, and nested loops over many rows.
At high volume: range-partition time-series tables by date so queries skip old partitions and archiving is cheap, batch the inserts, use keyset pagination instead of OFFSET, and remember that every extra index slows writes.

Follow-up

How would you speed up a slow join across several legacy tables with millions of rows each?
When would you denormalise, and how would you keep the copy consistent?
How do you check that a new index is actually used and is worth its write cost?

Keep soft-deleted accounts from blocking re-registration

Easy · soft delete · unique constraints · engine differences

app_user holds user_id, tenant_id, email CITEXT, password_hash (NULL for SSO principals), email_verified_at, auth_version, status ('invited','active','suspended','deactivated'), created_at, updated_at, deleted_at. Two live accounts for one address inside a tenant must be impossible, but an address freed by a soft delete must be reusable, and the same tenant may delete and re-register it repeatedly. Write the uniqueness DDL for PostgreSQL 16, then the equivalent for MySQL 8 where partial indexes do not exist, and say what each permits once three deleted rows already hold that address.

Approach

Start from what is actually unique: not (tenant_id, email), but (tenant_id, email) among live rows. PostgreSQL says that directly — CREATE UNIQUE INDEX app_user_live_email ON app_user (tenant_id, email) WHERE deleted_at IS NULL. A full constraint over the same two columns burns the address permanently the first time someone deletes an account.
Keep case-insensitivity in the type or the index, never in the application: CITEXT as given, or UNIQUE (tenant_id, lower(email)) as an expression index where the extension is unavailable. A case-sensitive unique column is exactly how two accounts for one human appear.
For MySQL 8 the predicate has to move inside the key: add a discriminator column that is a constant 0 while the row is live and is set to user_id on delete, with UNIQUE (tenant_id, email, deleted_marker). Live rows share the constant and still collide; deleted rows differ from each other and stop colliding.
State the NULL variant and its dependency: leaving the marker NULL for deleted rows also works, because a unique index treats NULLs as distinct — true in MySQL, and true in PostgreSQL only under the default NULLS DISTINCT, which PostgreSQL 15 lets you reverse. Check the polarity against the three existing deleted rows: constant-on-live is what preserves the collision you want, and reversing it silently admits duplicate live accounts.

Follow-up

A deleted account re-registers with the same address the next day. Do the old resource rows follow the new user_id, and how does the API keep the two principals apart?
How do you honour an erasure request while resource_revision.actor_user_id still references this table?

Walk me through the technical architecture of the most complex software system you designed or maintained in y

Medium · Technical Experience & Resume Deep Dive

Walk me through the technical architecture of the most complex software system you designed or maintained in your previous role.

Approach

Open with one sentence on what the system did and who used it, then walk the components in request order: clients, API layer, services, data stores, background jobs and external integrations. Say clearly which parts you built and which you inherited.
Give numbers you can defend: users, request rate, data volume, largest table size, latency targets. Say whether reads or writes dominated, because that explains most of the design.
Pick two decisions and give the trade-off for each, for example: "I kept reporting in a relational store because the queries were join-heavy, and I moved slow vendor calls onto a queue so user requests never waited on them."
Cover how it ran in production: deployment path, monitoring and alerts, and one incident with the change it led to.
Close with what you would change now and why. That shows you understand where the design stops working.

Follow-up

Which component fails first if load grows tenfold, and which metric would show it?
Why that data store and not the obvious alternative?
What happens to in-flight work when the queue or worker restarts?

What specific programming languages, frameworks, and data platforms did you use in your last project, and why

Medium · Technical Experience & Resume Deep Dive

What specific programming languages, frameworks, and data platforms did you use in your last project, and why were they selected?

Approach

List the stack by layer in one pass: language and runtime, web framework, data store, messaging, hosting, CI/CD. Then spend your time justifying the two or three choices that actually mattered.
Tie each reason to a constraint, not a preference. Examples: an existing enterprise standard (.NET or Java), team skills, a library ecosystem (pandas for data shaping), static typing for a long-lived codebase (C#, Java, TypeScript), or where the company already hosts (AWS or Azure).
If a choice was inherited, say so plainly, then say whether you would keep it and why.
Name a cost you lived with and how you handled it, for example ORM convenience causing N+1 queries that I fixed with eager loading or a hand-written join.
Be ready to go one level down on any item: the version, how it was configured, and a limitation you hit.

Follow-up

What would you pick for the same project if you started today?
What did that framework make harder than it should have been?
How did you compare alternatives before committing?

How have you integrated external APIs or third-party enterprise platforms (such as ServiceNow or custom ERP sy

Medium · Technical Experience & Resume Deep Dive

How have you integrated external APIs or third-party enterprise platforms (such as ServiceNow or custom ERP systems) into an existing technology stack?

Approach

Describe the integration's shape first: which direction data flows, sync or async, what triggers it (webhook, polling, scheduled batch), and which system owns each field. For ServiceNow, the REST Table API (/api/now/table/{table}) is the usual entry point.
Cover access and contract details: OAuth 2.0 client credentials or API keys kept in a secret store, rate limits, pagination, and a mapping layer that turns the vendor's schema into your own model so vendor changes stay in one place.
Make it reliable: timeouts on every call, retries with exponential backoff and jitter only where the operation is idempotent (upsert on the external record id, or an idempotency key), a dead-letter queue for records that keep failing, and a reconciliation job that compares counts or checksums.
Make it observable: correlation ids in logs, a stored last-synced watermark, and alerts on sync lag and error rate.
Test against the vendor sandbox or a mock, add contract tests, and say how you handled a version change or a deprecated field.

Follow-up

The external system is down for a day. What happens to changes made in the meantime?
A call times out and you retry. How do you avoid creating a duplicate record?
How did you handle a field-mapping change on the vendor side?

How would you design a data pipeline to ingest, validate, and visualize real-time operational data from remote

Medium · Software Design, Architecture & Systems

How would you design a data pipeline to ingest, validate, and visualize real-time operational data from remote project construction sites?

Approach

Clarify before designing: the data sources (sensors, equipment telemetry, field forms), event rates, how fresh the dashboard must be (seconds or minutes), and how reliable site connectivity is.
At the edge: a gateway at each site buffers to local disk and forwards with at-least-once delivery, batched and compressed. Every record carries site id, device id, event timestamp and a per-device sequence number.
Ingest through a partitioned log (Kafka, Azure Event Hubs or Kinesis) keyed by device id so each device's order is kept. A stream processor validates schema (JSON Schema or Avro), ranges and units, and sends rejects to a quarantine topic with the reason instead of dropping them.
Storage: keep raw data in immutable object storage for replay. Write cleaned data to a time-series store or date-partitioned tables, deduplicated on (device id, sequence) because delivery is at-least-once. Aggregate on event time with a watermark for late arrivals, and serve the dashboards from per-minute or per-hour rollups.
Operations: track freshness per site (last record seen) and alert on gaps. Backfill rollups from the raw zone when validation rules change.

Follow-up

A site is offline for two days and then uploads everything at once. What happens downstream?
A sensor starts reporting in the wrong units. How do you detect it and fix the history?
Why partition by device rather than by time?

Explain the key differences between monolithic and microservice architectures, and discuss when you would reco

Medium · Software Design, Architecture & Systems

Explain the key differences between monolithic and microservice architectures, and discuss when you would recommend each for internal enterprise tools.

Approach

Define both. A monolith is one deployable unit, usually with one database, in-process calls and simple ACID transactions. Microservices are independently deployable services that each own their data and talk over the network (HTTP, gRPC or messaging).
Give the cost side of microservices: network latency and partial failure, sagas and eventual consistency instead of one transaction, per-service CI/CD, distributed tracing, and container orchestration. Then the benefits: independent deploys and scaling, fault isolation, and separate ownership for separate teams.
For an internal enterprise tool with a modest user base and one owning team, recommend a modular monolith with enforced module boundaries. Extract a service only when one part has a different scaling profile, release cadence or owner, for example heavy report generation or a vendor integration.
For moving off an existing monolith, describe the strangler fig pattern: put a facade in front, route one capability at a time to the new service, and move that capability's data ownership with it.
Name team structure (Conway's law) and operational maturity as deciding factors, because a team without automated deploys and monitoring pays the microservice costs without getting the benefits.

Follow-up

How would you split a monolith whose modules all share one database?
How do you keep two services consistent without a distributed transaction?
What would make you merge two services back together?

What design patterns do you rely on most frequently when building scalable applications, and how do you apply

Medium · Software Design, Architecture & Systems

What design patterns do you rely on most frequently when building scalable applications, and how do you apply them?

Approach

Choose three or four patterns you have actually used and give a concrete case for each instead of reciting the catalogue. Strategy: swappable calculation or export rules. Adapter: a vendor API behind your own interface. Repository: data access isolated so logic can be unit-tested. Observer or pub/sub: change notifications.
Treat dependency injection as the enabler: constructor injection, built into ASP.NET Core and Spring, lets you swap an Adapter or Repository for a fake in tests.
For scale, name the architectural patterns: stateless services behind a load balancer, cache-aside with a TTL and explicit invalidation, queue-based load levelling for bursty work, a circuit breaker with retry and backoff around flaky dependencies, and CQRS when the read and write models really differ.
State the trade-off: every pattern adds indirection, so apply one where variation already exists, not speculatively. A Singleton holding mutable state is a concurrency and testing hazard.

Follow-up

Tell me about a time a pattern made the code worse.
How does a circuit breaker decide to let traffic through again? (A half-open state that sends trial requests.)
How would you unit-test the code that sits behind your Adapter?

How would you structure an automated reporting system that aggregates data from multiple engineering disciplin

Medium · Software Design, Architecture & Systems

How would you structure an automated reporting system that aggregates data from multiple engineering disciplines into a unified dashboard?

Approach

Clarify the consumers, the refresh cadence (daily or near real time), the exact metric definitions, and which system is the source of truth for each discipline's data.
Pull from each source with a connector (API, database replication or file drop) into a raw landing area. Transform into a conformed star schema with shared dimensions (project, area or work package, discipline, date) and fact tables such as progress, hours or quantities. Schedule it with an orchestrator such as Airflow or Azure Data Factory.
Load incrementally using updated_at watermarks or change data capture, and make every load idempotent (MERGE or upsert on the business key) so a rerun never double-counts.
Gate publication on data-quality checks: row counts against the source, null and referential-integrity checks, and unit consistency. Show each source's last refresh time on the dashboard so stale data is visible.
Serve the BI tool from aggregated tables, apply row-level security per project, and keep metric definitions in one documented semantic layer so two reports cannot compute progress differently.

Follow-up

Two disciplines code the same area differently. How do you reconcile them?
One source fails to load overnight. What does the dashboard show the next morning?
How do you handle a correction to last month's figures?

You are tasked with replacing a manual spreadsheet workflow used by field engineers with an automated web appl

Medium · Practical Problem-Solving & Case Scenari

You are tasked with replacing a manual spreadsheet workflow used by field engineers with an automated web application. How would you collect requirements and design the migration plan?

Approach

Start by watching field engineers use the spreadsheet and by collecting real copies. Inventory the columns, formulas, macros, validation rules, who edits what, and how files get shared, because the hidden logic lives in those files.
Turn that into user stories with acceptance criteria per role, and pick a first release that replaces the most painful step. Confirm every calculation rule with the domain experts and turn the spreadsheet's existing outputs into test cases with known expected values.
Design a relational schema in place of free-form cells, with server-side validation, an audit trail of who changed what, and role-based permissions. Check whether the app has to work offline at sites, and keep an Excel export because people will still need it.
Migrate with a repeatable import script that profiles and cleans the legacy data (duplicates, merged cells, text in numeric columns), then reconcile totals between the spreadsheet and the app. Pilot with one team, run both in parallel for a period, then make the spreadsheet read-only, with a rollback plan.
Support adoption with training and a feedback channel, and measure the outcome: time per entry, error or rework rate, and how many people are still using the spreadsheet.

Follow-up

People keep using the spreadsheet after launch. What do you do?
You find a spreadsheet formula that has always been wrong. Do you copy it into the app?
How do you import files whose column layouts differ between sites?

How do you balance writing pristine, fully tested code against delivering rapid solutions when field operation

Medium · Practical Problem-Solving & Case Scenari

How do you balance writing pristine, fully tested code against delivering rapid solutions when field operations demand immediate technical fixes?

Approach

Classify the issue by impact and reversibility. Blocked field work, safety or data-integrity problems justify a hotfix; cosmetic or workaround-able issues wait for the normal release.
State what you never skip, even in a hotfix: reproduce the bug, make the smallest change, add a regression test for that specific bug, get one other person to review it, ship through the normal pipeline, and keep a rollback path or feature flag ready.
Take extra care with irreversible changes such as data migrations or bulk updates: back up or snapshot first and test on a copy, because a fast fix there cannot be undone.
Afterwards, log the shortcut as tech debt with an owner and a date, do the proper fix with broader tests, and run a blameless review of why the bug reached the field.
Back it up with a real example in STAR form: the pressure, what you cut and what you refused to cut, and the outcome.

Follow-up

Tell me about a time a quick fix caused a second problem.
How do you stop tech-debt tickets from being ignored?
Who decides that something counts as an emergency?

Describe how you would implement user access control and data security policies for sensitive project document

Medium · Practical Problem-Solving & Case Scenari

Describe how you would implement user access control and data security policies for sensitive project documentation.

Approach

Start by classifying the documents (internal, confidential, restricted or export-controlled) and working out who needs which access. Apply least privilege and deny by default.
Model access as RBAC with roles scoped per project, plus attribute checks (ABAC) for classification, contract or clearance. Enforce it on the server for every request and at the object level, so changing a document id in a URL cannot expose another document (IDOR).
Authenticate through SSO (SAML or OIDC) against the corporate identity provider, with MFA, group-based role mapping, short-lived tokens, and narrowly scoped service accounts for integrations.
Protect the data itself: TLS in transit, encryption at rest with managed keys (KMS), short-lived signed URLs for downloads, and DLP or watermarking for the most sensitive classes.
Cover audit and lifecycle: an append-only access log, alerts on unusual bulk downloads, periodic access reviews, automatic removal when someone leaves a project (joiner-mover-leaver), retention rules, and automated authorisation tests per role.

Follow-up

An engineer moves from one project to another. What changes, and how quickly?
How do you make search work without leaking the titles of documents a user cannot open?
How would you detect and respond to a bulk download?

How do you ensure data consistency and system reliability across distributed engineering environments with int

Medium · Software Design, Architecture & Systems

How do you ensure data consistency and system reliability across distributed engineering environments with intermittent connectivity?

Approach

Frame it with CAP: during a network partition you either accept local writes (availability) or refuse them (consistency). Field tools usually go offline-first with eventual consistency, but name the data that must stay server-authoritative, such as approvals or unique numbering, and block those actions while offline.
On the client, write changes to a local store (SQLite or IndexedDB) and append each operation to an outbox with a client-generated UUID, so replaying after a dropped connection is idempotent. Sync deltas on reconnect using a server-issued version or change token.
Handle conflicts with optimistic concurrency: a row version or ETag sent in If-Match, and a 412 Precondition Failed on mismatch. Then choose a policy per data type: last-writer-wins (simple, but it silently loses edits and depends on clock accuracy), field-level merge, or showing the conflict to the user. CRDTs fit counters and sets that must merge automatically.
On the server, use the transactional outbox pattern so the database write and the published event cannot diverge, and make consumers idempotent. At-least-once delivery plus deduplication gives effectively-once processing.
Record deletes as tombstones so they sync. Show sync status to the user, monitor sync lag and conflict rate, and test with dropped and partial uploads.

Follow-up

Two engineers edit the same record while offline. What does each see after syncing?
Why not just use last-writer-wins with device timestamps?
How do you sync a delete made offline?

If a critical software tool used by field teams experiences unexpected latency during peak operating hours, ho

Medium · Practical Problem-Solving & Case Scenari

If a critical software tool used by field teams experiences unexpected latency during peak operating hours, how would you systematically investigate the root cause?

Approach

Scope the problem first: which operations, which sites or users, since when, and whether it tracks load or a deploy. Measure p50, p95 and p99 instead of averages, and compare server-side timings with what clients see, since remote sites may have slow network links.
Follow the request path (client, network or VPN, load balancer, app servers, database, external services) and use distributed tracing or timing logs to find the span that grows at peak.
Check saturation with the USE method (utilisation, saturation, errors) on CPU, memory and GC pauses, thread pools, database connection-pool waits, lock blocking and the slow-query log, plus rate limits on any external API.
Because it only happens at peak, look for queueing: latency climbs steeply as utilisation nears 100%. Also check for batch jobs, reports or backups scheduled into busy hours, and for lock contention on hot rows.
Mitigate (scale out, move the batch job, add the missing index, cache), reproduce the peak with a load test to confirm the cause, verify p95 has recovered, and add a dashboard and alert for that signal.

Follow-up

Server metrics look normal but users still report slowness. What next?
How do you tell connection-pool exhaustion apart from slow queries?
How would you build a load test that reproduces the peak pattern?

Suggested practice checklist

Each day targets a reported question group or round: resume and integration depth first, then design and case scenarios, then behavioral stories, ending with a full spoken run-through.

Day 1 — Resume inventory and screening answers

List every project on your resume with its stack, your role, one architecture decision and one regret.
Write spoken answers for the architecture walk-through and the languages, frameworks and data platforms questions, using two different projects.
Write down your position on relocation, site travel, clearance eligibility (if relevant) and pay range.

Deliverable: A project inventory table, two rehearsed project answers, and a short script for the HR screen logistics.

Day 2 — Integration and data layer

Outline a ServiceNow or ERP integration: authentication, mapping, retries, idempotency, error handling and test strategy.
Take a slow query across several large tables: write the schema grain, indexes and an EXPLAIN-based tuning sequence.

Deliverable: An integration outline, a tuned-query walkthrough, and both practice drills completed.

Day 3 — Pipelines, offline sync and dashboards

Draw a pipeline from remote site sensors or forms to a dashboard: ingestion, validation, storage, alerting, visualisation.
Write a short design for syncing data from sites with intermittent connectivity: local queue, ordering, conflict resolution and resync.

Deliverable: Three diagrams (pipeline, offline sync, reporting service) with a failure-mode table for each, plus the completed interval drill.

Day 4 — Architecture trade-offs and migration

Write a decision table for monolith versus microservices for an internal enterprise tool: team size, deployment, data coupling, operational cost.
Prepare three design patterns you actually used (for example repository, strategy, adapter), each with a concrete example from your code.
Outline requirements gathering and a staged migration from a field spreadsheet to a web app, including parallel running and rollback.

Deliverable: A decision table, three pattern examples, and a migration plan with phases and exit criteria.

Day 5 — Diagnosis, delivery pressure and access control

Write a latency runbook: what changed, what to measure, how to isolate each layer, and how to confirm the fix.
Prepare a position on quality versus speed for urgent field fixes: hotfix criteria, minimum tests, follow-up refactor.
Design role-based access for sensitive project documents: roles, permissions, audit logs, encryption and review of third-party files.

Deliverable: A one-page latency runbook, a hotfix policy outline, and an access-control model with an audit plan.

Day 6 — Behavioral story bank

Write six STAR stories, one per reported behavioral prompt, each with one decision of your own and a quantified result.
Add a technical detail to each story (system, cause, fix) in case the round turns specific.
Say each story aloud and cut it to the version that makes one clear answer; test one with a non-software friend.

Deliverable: Six written STAR stories with a technical detail line each, rehearsed aloud once.

Day 7 — Full run-through

Run a mock loop aloud: a short career summary, one resume deep dive, one design question and two behavioral stories.
Review gaps from days 1 to 6 and rewrite the weakest answer.
Prepare questions for the panel about the stack, modernisation work and how the software team works with field operations.

Deliverable: A recorded or written mock run, one rewritten answer, and a list of panel questions.

Stories for the Bechtel behavioral round

Candidates describe the last round as a mix of behavioral and technical questions that probes interpersonal skills. The reported themes fit a role between software and project teams: working with non-technical stakeholders, conflicting requirements, deadlines and scope changes, owning mistakes, disagreeing with senior colleagues, and adapting to new domains or locations. Reports also mention handling a stakeholder unhappy with a delivered tool and influencing a decision without direct authority, so prepare those angles too. Prepare each story in STAR form and keep a technical detail ready.

Describe a time when you had to debug a critical issue in production or during a live project deployment. How

Medium · Technical Experience & Resume Deep Dive

Describe a time when you had to debug a critical issue in production or during a live project deployment. How did you diagnose and resolve it?

Approach

Choose an incident where you drove the diagnosis. Open with the symptom and its blast radius in two sentences: who was affected, which workflow stopped, how you knew.
Show the diagnostic sequence, not a lucky guess: what changed recently (release, config, data), what logs and metrics you checked, how you narrowed by layer (client, API, database, integration), and how you reproduced it.
Separate stabilising from fixing. Say whether you rolled back, applied a feature flag or patched forward and why, then describe the root-cause fix and the regression test or alert you added.
Close with the outcome and the process change, such as a runbook, a deployment check, or a staging rehearsal for live deployments. The mistake that sinks this story is a team narrative where your own decisions are missing.

Follow-up

Why did you choose a rollback instead of a forward fix, and what would have changed that choice?
How did you confirm the root cause rather than a coincidence with the release?
What would have caught this before it reached users?

Describe a situation where you had to work with non-technical stakeholders or domain experts who had conflicti

Medium · Behavioral & Situational Judgment

Describe a situation where you had to work with non-technical stakeholders or domain experts who had conflicting requirements. How did you align them?

Approach

Name the stakeholders and the conflict concretely: for example, two discipline groups wanting incompatible fields, approval flows or report layouts. State what each party needed underneath the stated position.
Describe how you made the conflict visible: a side-by-side requirements table, sample scenarios, or a clickable mockup that let each side react to the same thing.
Frame the trade-offs in their language (schedule, rework, cost, safety effect), then show how the decision was made: who had authority, what you recommended and why.
Finish with a written outcome that everyone confirmed, what shipped, and how conflicts were checked later. The story fails if you only relayed messages or quietly picked a winner.

Follow-up

What did you do when neither side would move, and who made the final call?
How did you record the agreement so it did not reopen later?
What did you do when a conflicting requirement surfaced after development began?

Tell me about a time you faced tight project deadlines or unexpected scope changes. How did you prioritize you

Medium · Behavioral & Situational Judgment

Tell me about a time you faced tight project deadlines or unexpected scope changes. How did you prioritize your workload?

Approach

State the deadline and what changed in one or two sentences, such as a late feature request from a department head before a release. Then show how you re-ranked work: must-have versus deferrable, user impact versus effort, dependencies that could block.
Show you negotiated scope instead of silently absorbing it. Present options to the requester (cut, defer, add time or help) with the consequence of each, and say which one was agreed.
Protect the risky path: keep tests and review on the parts that could break live use, and cut polish or low-use features first. Mention how you told affected people what was left out.
Report what shipped, what was deferred and when it was picked up. Working long hours is not the answer; the story needs your prioritisation logic.

Follow-up

What exactly did you cut, and how did you decide it was safe to cut?
How did you tell the requester no, or not yet?
What happened to the deferred items afterwards?

Give an example of a mistake you made on a technical project. What steps did you take to fix it, and what did

Medium · Behavioral & Situational Judgment

Give an example of a mistake you made on a technical project. What steps did you take to fix it, and what did you learn?

Approach

Choose a real mistake you owned, with enough consequence to matter but not a disaster you cannot explain, for example a faulty migration step, a missing validation, or a mis-scoped release. Say plainly what you did wrong.
Describe containment first: who you told and how fast, how you limited damage (rollback, data restore, user notice), then the root cause stated in process terms such as a missing test, review step or environment gap.
Show the repair and the guard you added: an automated check, a pre-deployment checklist, a dry run, monitoring. Include the change in your own habits.
Avoid the humble-brag mistake (working too hard) and blame-shifting. The story works when the lesson is specific and checkable.

Follow-up

How soon did you tell your lead, and what did you say?
Which check would have caught this, and is it in place now?
Knowing what you knew at the time, would you make the same call again?

How do you handle situations where you disagree with a technical decision made by a senior colleague or team l

Medium · Behavioral & Situational Judgment

How do you handle situations where you disagree with a technical decision made by a senior colleague or team lead?

Approach

Pick a specific technical decision, such as a data model, framework or release approach. Explain the senior person's reasoning fairly before saying why you disagreed, because it shows you understood the constraint.
Describe how you raised it: privately first, with evidence (a small spike, a benchmark, a written comparison of options and risks), and phrased as options rather than as a verdict.
Say how it resolved. Either the decision changed because of the evidence, or you committed to their direction and helped it succeed, noting what you tracked to see who was right.
The story fails if it is only about you being right or if you went around the person. It should show respect, evidence and a clean handoff to the decision-maker.

Follow-up

What would you have done if they still disagreed after seeing your data?
Tell me about a time you turned out to be wrong in a disagreement.
When would you escalate a technical disagreement, and to whom?

Describe a time when you had to adapt to working on a project in an unfamiliar technical domain or geographic

Medium · Behavioral & Situational Judgment

Describe a time when you had to adapt to working on a project in an unfamiliar technical domain or geographic location.

Approach

Say what was unfamiliar: a business domain, a hardware or process-engineering concept, a different tech stack, or a different country, site or time zone. Be specific about what you did not know on day one.
Show the ramp-up plan: shadowing users, interviewing domain experts, building a glossary, reading their existing documents, and asking for early review of your first small change.
Explain how you adapted communication to the setting, for example confirming requirements in writing across time zones or using diagrams for non-software colleagues.
Give a concrete result: what you delivered, what domain experts relied on you for, and what you would do faster the next time. Enthusiasm without a ramp-up method is the weak version of this story.

Follow-up

What did you get wrong in the first weeks, and who corrected you?
How did you decide whom to ask and whom to read?
Would you take a site assignment or relocate for a project, and how would you prepare?