10 D. E. Shaw Quantitative Analyst (New Grad) Interview Questions (2026)

Outline

STAR. Pick a real problem — coursework or research — where the initial approach failed. Walk through how you reframed the problem, what insight unlocked progress, what the final solution looked like. D. E. Shaw is testing whether you have genuine grit and creative problem-solving, not pattern matching.

Outline

Specific reasons: multi-strategy with both systematic and discretionary groups, deep research culture, computational biology and venture arms show intellectual breadth. Reference the firm's history (founded 1988, early systematic trading). Avoid the 'high pay' lead. Show you have read at least one D. E. Shaw research paper or blog post.

Outline

Kadane's algorithm. Track max_ending_here = max(num, max_ending_here + num) and max_so_far = max(max_so_far, max_ending_here). O(n) time, O(1) space. Edge case: all negative numbers — return the largest single element. Walk through with a small example; D. E. Shaw values clean code over speed.

Outline

Two heaps: a max-heap for the lower half, a min-heap for the upper half. Add: insert into max-heap, then rebalance — pop max of lower, push to upper if needed, and ensure size difference ≤ 1. Median: if equal sizes, average tops; else top of larger heap. O(log n) add, O(1) median. Walk through with an example.

Outline

Unfold the cube into a 2D net. The shortest path is a straight line across two adjacent faces. Distance = sqrt(1^2 + 2^2) = sqrt(5) ≈ 2.236. Discuss why unfolding works (preserves distances along faces). Be ready for follow-up: ant on a rectangular box with different dimensions — same trick but check which unfolding gives the shortest path.

Outline

Bertrand's paradox. Answer depends on how you sample. (1) Pick random endpoint angles: P = 1/3. (2) Pick random midpoint along a fixed radius: P = 1/2. (3) Pick random midpoint uniformly in the disk: P = 1/4. State all three and explain the dependence on sampling method. D. E. Shaw tests whether you recognize ambiguity, not whether you compute one specific answer.

Outline

Classic counterfeit coin puzzle. First weighing: split 4 vs 4, set 4 aside. If balanced, odd is in the set aside. If not, narrow to the 8 weighed balls. Second and third weighings use known-good balls as references to triangulate. Walk through one branch carefully — the recruiter wants to see the search-tree thinking, not the full enumeration.

Outline

Gaussian integral. Square the integral, convert to 2D, transform to polar coordinates: (∫e^(-x^2)dx)^2 = ∫∫e^(-(x^2+y^2))dxdy = ∫_0^{2π}∫_0^∞ e^(-r^2) r dr dθ = π. So the original integral = sqrt(π). State this is the foundation of the normal distribution's normalization constant.

Outline

Sneaky simple answer. Trains close at 100 mph combined, so they meet after 1 hour. Bird flies for 1 hour at 75 mph, so it travels 75 miles. Don't try to sum the infinite series of bird back-and-forths — that's the trap. State the trick upfront, then the answer.

Outline

Classic random-walk problem. 1D symmetric random walk is recurrent — probability of reaching any integer is 1. Derive: let p = P(reach +1 from 0). From 0, first step is either +1 (probability 1/2, done) or -1 (probability 1/2, need to walk back to 0 then to +1, each with probability p). So p = 1/2 + (1/2)p^2. Solve: p = 1 (the other root p=1 is unique valid). Discuss why this fails in 2D and 3D.