Divyansh Lalotra nej1gotnochill

Convex Optimization Engineer. Building mathematical software that measurably runs faster.

I work where mathematical software meets the machine — canonicalization pipelines, solver interfaces, and the benchmarking infrastructure that tells maintainers whether a change actually made things faster.

Open Source, Engineering & Programs

cvxpy Open Source Developer — added dual variables to ConstantSolver for feasible constant problems (#3505, merged) and isolated OR-Tools tests from HiGHS to prevent native symbol conflicts (#3508, merged)
cvxpy Tuple-Axis Norms — tuple-axis support in norm() plus an N-D p=2 SOC canonicalization fix, including a per-fiber coordinate-mapping bug that keepdims=True exposed while value-level assertions stayed tautological (#3539)
OpenMS Open Source Developer — built the project's first benchmarking layer for Issue #8788: a gated ENABLE_BENCHMARK_TESTING CMake option, benchmark test directory, and CTest registration (#9839)
OpenMS-benchmarking Benchmarking Infrastructure — CI that builds OpenMS from a pinned SHA, runs DDA and OpenSwath DIA benchmarks, and reports wall time, CPU time, and peak RSS against a stored baseline. Contributed the v2 results schema with automatic v1 promotion, the benchmark-agnostic report renderer, the OpenSwath DIA benchmark and real comparison baseline, package-mode runtime, and the nightly benchmark loop (#3, #4, #5)
Netra ML Security — passive network threat-detection platform: a leakage-aware DDoS detector validated under a strictly capture-disjoint protocol (Precision 0.999998, Recall 0.997849, FPR 0.00018 on 448k unseen flows) plus a streaming multi-threat service with causal early-detection windows
cyberworld Predictive SOC — turns a SPAN feed into a live host graph with ATT&CK-aware risk forecasting, built on PyTorch temporal-graph models (TGN) for attack-evolution prediction
HWSS-ML Research — intelligent authentication research combining ensemble risk scoring with SHAP explainability, focused on adversarial robustness and concept drift; volatility forecasting work spanning HARNet, GARCH, and gradient-boosted models with sentiment features