Open-source AI systems engineer in Pune building AI harnesses, inference control planes, and developer workflow tooling.
I work best where system design, hands-on implementation, and maintainable open source meet.
Maintaining HarnessLab, building codex-stack, and open to strong engineering teams, early-stage products, and meaningful OSS collaboration.
- I build and maintain developer-facing AI systems in the open
- I care about contributor-friendly repositories: clear architecture, useful docs, strong tests, and practical APIs
- I am especially interested in AI infrastructure, harness runtimes, inference systems, and workflow tooling
- I ship full systems, not isolated demos: runtime code, APIs, CLIs, tests, docs, and repository structure
- My work tends to be infrastructure-shaped: harness layers, tracing, policy systems, evaluation flows, scheduling, and automation
- I am comfortable moving from architecture to implementation to polish without handing off the hard parts
- HarnessLab
Open-source platform for learning and simulating how real AI systems work internally, including agent harnesses, inference, memory, cost, and control layers. - codex-stack
Codex-native stack for review, browser QA, preview verification, and shipping automation. - awesome-codex-skills
Reusable Codex Skills for AI engineering workflows. - context-pack
Task-specific repo context bundler for AI coding agents.
- Strong fit for AI systems, developer infrastructure, and applied platform engineering roles
- Comfortable owning architecture, implementation, and technical quality end-to-end
- Strong TypeScript, Bun, Python, and repo-automation depth
- Best fit for early teams building AI infrastructure, developer tools, or workflow products
- I like 0→1 work where product shape and system design evolve together
- I optimize for leverage and shipping speed without lowering the engineering bar
- Open to collaborating on technically serious infrastructure projects
- Interested in maintainable abstractions, readable documentation, and contributor-friendly repo design
- Especially aligned with projects around AI systems, automation, developer tools, and code quality
- AI harness engineering systems: guardrails, tracing, memory, evals, policy layers, and approval workflows
- Inference control-plane simulations: prefill vs decode, KV cache, batching, latency, throughput, and cost modeling
- Agentic workflows for code review, browser QA, preview verification, and shipping readiness
- Repo-native developer tooling, reusable Codex skills, and ReviewOps automation
If you are hiring for AI systems work, building an early-stage AI tooling company, or want to collaborate on open-source infrastructure, reach out.