Hey, I’m
I build an information network for AI agents.
Agents are only as useful as what they can see and touch. At Caylex I build the layer that decides what an agent knows, which tools it can reach, and what it’s allowed to do.
Founding engineer at Caylex, building the infrastructure that connects AI agents to enterprise software over MCP. Before this I took a client's vertical-AI product 0→1 at TheAgentic, and led the AI observability work at Middleware Labs (YC W23) — log clustering, root-cause agents, and natural-language querying.
Own context engineering and agent infrastructure end-to-end. Built the MCP proxy — a unified, secure interface for agents to discover, authenticate with, and call MCP servers across hundreds of enterprise systems — plus the orchestration and governance around it.
Took a client's vertical-AI product 0→1 at a venture studio: frontend, backend APIs, GitHub/Jira integrations, and launch tooling — deep-reasoning agents and tool-calling pipelines on Python and TypeScript. Four months, shipped.
Led the AI observability work on a cloud platform. ML log clustering improved RCA speed by 60%, cutting incident resolution from hours to minutes; also shipped an MCP server streaming logs & traces to LLMs, anomaly detection, natural-language querying, and forecasting.
Schedulers for third-party integrations and a data pipeline for BloomChat AI; data quality via Pydantic models in FastAPI.
Led a team delivering backend APIs and cloud deployments; built computer-vision applications and data/processing pipelines.
Earlier: ML & computer-vision internships — Leading India AI (hyperspectral image reconstruction), Crear Electronics (Furnysh), and SMART Foundry (DST, Govt. of India), 2019–2020.
ML log-pattern clustering that improved RCA speed by 60%, cutting incident resolution from hours to minutes across large-scale cloud environments.
An AI agent that pinpoints failure sources across distributed systems and proposes concrete code & config fixes.
AI-powered observability: log anomaly detection for early issues, a natural-language query tool for SREs to ask about logs & infra in plain English, and time-series forecasting & outlier detection for sharper alerts.
AI-powered photo sharing & discovery. AI engine + backend on Node.js · PostgreSQL · Vector DB · AWS. Featured at Laracon India 2023.
Federated-learning platform. Cross-platform worker apps in Go & Python; chunked training in isolated environments synced to an averaging server.