Open to opportunities

Krish Goyani

@krishgoyani

Message

AI Agent Engineer building production-grade multi-agent systems for reliable code generation.

India

Message

What I'm looking for

I’m looking for a team where I can architect and ship production LLM/AI agent systems end-to-end—multi-agent reasoning, retrieval, evals, and observability—while driving cost and reliability. I thrive on architecture-first ownership and measurable quality.

I’m an AI Agent Engineer who builds multi-agent systems from first principles—reasoning loops, retrieval, evals, observability, and cost optimization. I own Agent v2 at Rocket.new, the core engine behind every code-generation task on the platform.

I architected a self-healing autonomous code-generation system with a 6-phase reasoning framework, cutting cost per task 38% (up to 60% on long sessions) and steps-to-solution 28%. I also built an embedded quality/eval layer to verify generated code against measurable rules, driving error rates down across type-system, dependency, code-structure, and integration checks.

I ship production LLM and AI agent systems in Python, Go, and TypeScript, keeping everything stack-agnostic and architecture-first. From a VS Code AI coding agent (Rocket Copilot) to privacy-first indexing and dynamic plug-in connectors powering 25+ integrations, I focus on systems that perform reliably in the real world.

Experience

Work history, roles, and key accomplishments

Current

AI Agent Engineer

Current

Rocket.new

Jan 2025 - Present (1 year 6 months)

Architected Agent v2 end-to-end, a self-healing autonomous code-generation system with a 6-phase reasoning framework. Built an eval/quality verification layer, connector plug-in architecture, a WordPress-to-Next.js redesign engine, and a VS Code AI coding agent with MCP tooling and human-in-the-loop approvals.

Multi Agent Systems Reasoning Loops RAG Observability Cost Optimization Python Go TypeScript Next.js Semantic Search Human In The Loop MongoDB Pinecone

AI Engineer

Codehawk Technologies

Jan 2024 - Dec 2024 (11 months)

Designed a supervisor-worker multi-agent system using Pydantic-validated structured outputs for inter-agent communication, backed by async-first FastAPI services. Built a multimodal (vision + text) document extraction pipeline with intent routing, and created Pandas/NumPy pipelines to turn production logs into evaluation datasets via RESTful APIs with rate limiting.