CSE Undergraduate at NIT Silchar (2024–2028) · Backend Engineering · Distributed Systems · Applied AI

Email · LinkedIn · Codeforces · LeetCode

• Backend systems and distributed infrastructure — sandboxed execution, worker-coordinator architectures, event-driven pipelines
• Applied AI/ML for real-world decision systems: risk scoring, recovery workflows, traffic and routing intelligence
• Low-latency systems and quantitative technology, alongside competitive programming and open source

Pratyānayana — AI agent for merchant revenue recovery An autonomous agent that detects payment failures and checkout abandonment, predicts recoverability, and makes bounded Retry/Wait/Switch/Stop decisions using customer history and expected-value economics, executed through event-driven, idempotent Razorpay webhooks with causal recovery attribution. Benchmarked on 5,000 journeys — recovered ₹3.65 Cr of ₹5.00 Cr at risk, ₹45.5L incremental lift. FastAPI CatBoost Razorpay APIs SQLite LLM

Vahini — distributed trading-system benchmarking platform (demo) A distributed platform for evaluating trading engines: automated correctness validation, a Go-based load-generation framework, and a sandboxed Docker + gVisor execution pipeline for untrusted participant code, orchestrated via FastAPI worker-coordinator architecture. Sustains 1.3K+ TPS under 1,000 concurrent traders. FastAPI Go Docker gVisor AWS EC2/RDS PostgreSQL

ParivahanSaathi — AI/ML traffic intelligence and routing (demo) A city-scale traffic intelligence system for Bengaluru: a multi-stage inference pipeline for incident-based closure prediction, impact scoring, and diversion routing, validated in real time against Mappls traffic-aware APIs. 8K+ incidents, 56 partitioned OSM graphs, 94% memory reduction (4.8 GB → 280 MB), sub-150 ms warm route generation. FastAPI CatBoost OSMnx NetworkX AWS Mappls APIs

• TitanDB — Bitcask-inspired KV store in C++ with O(1) hash-index lookups, disk-backed storage, log compaction, and crash recovery
• TrafficTalk — real-time network traffic analysis with an LLM-powered (Groq + LLaMA 3) hybrid agent for insights
• FXStatArb — quantitative FX engine
• CAPO — cross-asset portfolio optimization

Also working with: REST APIs, CatBoost, LLMs, SQLite, distributed systems, performance optimization, data structures & algorithms.

• Layer5 — ongoing contributions to documentation and UI
• Apothesis — merged PR resolving cross-platform build failures on Windows
• FOSSOLOGY (Atarashi) — accepted PR fixing runtime crashes in the TF-IDF agent via exception handling in Nirjas-based comment extraction

• Semifinalist, GridLock Hackathon (Flipkart × Bengaluru Traffic Police)
• Semifinalist, Tata InnoVent
• Top 2.1% of teams (solo), IICPC Summer Trading Hackathon 2026
• 2nd Prize, Arbitrage Arena 2026, Indian Institute of Science
• 3rd Position, IISc Honour Code Hackathon
• Rank 2,245, Meta HackerCup Round 1 2025

_{Building systems where correctness and latency matter more than lines of code.}