NAGRAJ GAONKAR.

Software Intern, May–July 2026. Built AI response caching with vector search and BM25, cutting latency by 15 seconds and saving 600k tokens/month.

Wildfire forecasting with PatchTST + DLA, tree models, and Meta-MLP stacking, plus seven-day Monte Carlo risk simulation.

C++ ray tracer using MPI + OpenMP across 40 CPUs. Reduced a 1600 × 900 render from 19.64 hours to 0.81 hours.

Added Clang constexpr support for X86 shuffle intrinsics across SSE, AVX2, and AVX512.

Synthetic-image detection with CNN + ViT ensembles, adversarial testing, and visual explanations.

Search across 1,400+ documents using LSA and K-Means. Improved mAP by 9.3% over the VSM baseline.

Multithreaded F1 simulator with 20 ms telemetry, a bounded ring buffer, and a live terminal leaderboard.

Regime-aware intraday strategies with parallel backtests and no-lookahead constraints. Built for InterIIT.