Founding Software Engineer (but also an intern)

Kello is an AI intelligence platform backed by Peak XV (formerly Sequoia India). I was one of its first engineers, part of a 4-man team, and worked across both halves of the product.

The AI layer. Two production LangGraph workflows — a multi-turn pipeline turning natural-language recruiter queries into structured Elasticsearch filter plans, and an 11-step onboarding wizard with interrupt-driven UI and parallel JD summarization. Multi-model orchestration across Azure OpenAI and Google Gemini, with Langfuse for prompt versioning and tracing, caching by prompt version, and real-time streaming with heartbeats for long-running calls.

The data layer. Mining signal from 850M+ profiles worldwide — pulling from GitHub, Product Hunt, Devfolio, Devpost, and LeetCode. The build I'm proudest of: a tool that takes any company's website and turns it into clean, structured data — products, services, metrics, awards, and more. The catch is that no two company sites share a structure, yet one pipeline normalizes all of them into a single common schema.

I worked as an AI builder (agents, frameworks, MCP) and an AI-native builder — using AI to ship faster and going deep on the agentic-dev ecosystem in the process.