Two Sigma Quantitative Software Engineer Interview Experience — Order-Matching HackerRank Then a Quant Systems Deep Dive

Round 1: HackerRank Coding — Stock Exchange / Order Matching

The task was to implement buy/sell order matching:

• Buy side uses max priority.
• Sell side uses min priority.
• Match orders when the price can cross.
• Handle partial fills.
• Output trade records.
• Handle the remaining resting orders.

Follow-ups:

• How to add more tests to increase confidence.
• Why my output didn't match the sample.
• How multiple orders at the same price should be ordered.
• Should you use order ID, quantity, or arrival time for tie-breaking.
• Why choose a heap.
• Could you use a linked list, tree, or some other structure instead.
• Is a heap really optimal.
• What would you change with more time.
• How to improve the final string concatenation and result aggregation.

The core of it wasn't just getting the happy path to run — it was really testing:

• price-time priority;
• deterministic ordering;
• data structure trade-offs;
• testing/debugging discipline.

Round 2: Experience / Quant Systems Deep Dive

The second interviewer, a senior one, first had me walk through my resume completely, then had me pick a genuinely complex problem to go deep on.

Tax-loss harvesting optimizer

• Tax regions and binary variables.
• Big-M.
• MIQP.
• Whether it's essentially branch-and-bound underneath.
• Whether to do a continuous relaxation first.
• Whether to use the relaxed solution to shrink the binary search space.
• How short-term gain enters the utility function.
• Whether to consider waiting for the holding period to convert to long-term.
• Wash-sale constraints.
• The difference between single-period and multi-period.

Multi-period asset-location backtest

• Whether Monte Carlo paths are used.
• Rebalancing cadence.
• What metrics are being optimized.
• How much historical data is used.
• What data sources are used.
• How to get through 30,000 optimizer runs.
• Single machine or multiple machines.
• Whether running each path on its own machine would be faster.
• Shared inputs, serialization, and scheduling overhead.
• The actual runtime per path.
• How to profile the bottleneck.
• Which enhancements didn't get done in time.

Role fit

Then at the end they asked directly:

• Do you lean more toward pure software engineering or pure quant research?
• Which side are you strongest at and most interested in?
• Are you suited to the intersection of engineering and optimization research?