About
I chase the layers others treat as magic.
The fastest way I understand a system is to rebuild it — so I keep building the parts most people take for granted.
Software engineer drawn to the systems others treat as magic — a UI runtime, a language interpreter, a container runtime and a distributed datastore, all built from scratch. The fastest way I understand a system is to rebuild it.
I'm currently an Agentic AI Engineer at Zimyo, building LangChain/LangGraph agents that turn natural language into multi-step HR workflows — recently driving LLM inference cost down ~70% with schema-constrained prompt design. Before that, at Arrowhead Capital, I built a vectorized backtesting engine over 5M+ records and the pipelines behind it.
I'm a named inventor on a published Indian patent for a multilingual RAG system over the Bhagavad Gita, and an open-source author with 1,000+ npm downloads and 167 GitHub stars. I care about correctness, performance and clean architecture — and I'm always happy to talk developer tools, infrastructure and building from first principles.
What I work on
Distributed Systems
Sharding, replication and fault tolerance — key-value stores and load balancers that stay up when nodes don't.
Systems Programming
Low-level work in Go, Rust and C — event loops, interpreters and SIMD-vectorized engines built for speed.
Applied AI
Production agentic systems on LangChain/LangGraph and RAG pipelines — measured on cost and retrieval accuracy.
The track record
Where it shipped.
Education
VIT Vellore
2024 — 2026MCA — Master of Computer Applications
GPA 9.5 / 10
PIMR, Gwalior
2021 — 2024BCA — Bachelor of Computer Applications
GPA 9.6 / 10 · Gold Medalist
R.S.B. Inter High School
2019 — 2021Higher Secondary (12th), Science — BSEB
74%
R.S.B. Inter High School
2017 — 2019Matriculation (10th) — BSEB
84%
Highlights
- MCA, VIT Vellore — GPA 9.5/10
- Graduation Gold Medalist
- Named inventor · published Indian patent
- Microsoft Learn Student Ambassador
Agentic AI Engineer
ZimyoGurugram
- Built an agentic AI platform for HRMS automation on LangChain + LangGraph — natural-language multi-step workflows across payroll, attendance and employee-lifecycle modules.
- Engineered a dynamic API-ingestion engine that parses REST schemas and auto-generates executable workflow graphs, killing manual config for 5+ HR modules.
- Cut LLM inference cost 70% by consolidating multi-turn generation into a single schema-constrained call.
Quantitative Developer Intern
Arrowhead Capital ManagementMumbai
- Designed a vectorized backtesting engine over 5M+ time-series records using SIMD intrinsics — 18s → 4.2s (4.3× speedup).
- Built ETL pipelines with multi-layer validation + caching (+30% research throughput) and rolling-window risk analytics (Sharpe, max drawdown, VaR).
Full-Stack Developer Intern
Chiti InfotechGwalior
- Migrated a legacy PHP monolith to a modular React SPA (40% faster frontend) and reworked the REST layer with caching (−25% latency).
Open to hard problems
Got a system worth
building right?
Distributed backends, AI agents, or performance work that needs to go fast. Let's talk.