I build systems that are meant to be fast, resilient, and hard to break — distributed infrastructure, low-latency backends, and AI tooling. Currently at Quadeye building tools for quant workflows and data processing. CS Engineering graduate from IIIT-Delhi.
InferMesh · Distributed AI Inference Scheduler
Open-source scheduler that routes LLM requests across a cluster of self-hosted nodes in real time.
- Weighted scoring across latency, queue depth, and active requests — lowest score wins
- UDP gossip protocol for zero-config peer discovery and automatic eviction
- Transparent failover — if a node dies mid-request, the next best node takes over
- OpenRouter as a configurable remote spillover peer
- Full Prometheus metrics per node, OpenAI-compatible API surface
Python FastAPI Ollama Prometheus
Raft Algorithm · Consensus from Scratch
Complete implementation of the Raft distributed consensus protocol in Python.
- Leader election, log replication, and snapshotting — all from scratch
- 99.9% uptime in cluster fault simulations
- Handles node failures, network partitions, and live cluster expansion
Python Distributed Systems Consensus
Mimir · [description based on your project]
[Add a one-line description — let me know what Mimir does and I'll sharpen this]
stack stack
TradeMind · Autonomous Trading Agent
LLM-powered trading agent that ingests live market data and executes strategy autonomously.
- 19.6% ROI over 3 months on paper trading
- Custom tool-use and prompt architecture built on LangChain
- Integrated with 5+ real-time market data APIs
Python LangChain LLMs
Always building something. Open to interesting problems.
