Ishaan Sandhwar

Every GATE DA question from 2024–2026, taken only from the official PDFs and tagged to the official syllabus: weightage, a backtested forecast, an adaptive study planner and a web atlas.

My GitHub profile: AI/ML engineering projects with verified results and honest team roles, plus milestones and a self-hosted GitHub stats card.

Expands a part number and a short distributor description (median 35 characters) into a validated 252-column product record, with evidence, method and confidence on every value. Deterministic decoders first: 71% classified with no LLM call, 1,000 rows in about 7 s.

UPI fraud-ring detection & merchant risk analytics. TransOrg AgentIQ Datathon, Track 1 — Best Pipeline Award. Team of 5; my part: data engineering and the analytics model.

Disaster response simulator over a 40-node city graph: Dijkstra/A* routing, BFS flood spread, Edmonds–Karp evacuation max-flow, Tarjan bridge detection. Heap, hash map, trie and union–find written from scratch in C++17; 96 automated checks.

Binary PSO feature selection on the Kaggle credit-card fraud data: 30 → 7 features with no detectable loss against the same random forest on all 30 (ROC-AUC 0.977 vs 0.986; 98 test frauds, bootstrap CIs reported).