CI/CD | PyTorch Contributor | India

I work on CI/CD infrastructure and developer tooling for machine learning frameworks. Currently building the Cross-Repository CI Relay (CRCR) for PyTorch — a fully automated pipeline that triggers and tracks CI across downstream repositories whenever a PR is opened on pytorch/pytorch, with results flowing back to the PyTorch CI HUD.

• Cross-Repository CI Relay (CRCR) — Designed and built the relay system that connects upstream PyTorch events to downstream CI across the ecosystem. Key components: Callback Lambda (OIDC verification, state machine, rate limiting), HUD integration (React dashboard pages, ClickHouse queries, API routes), on-call bot, and CI-neutral callback action (GitHub Actions + Buildkite). Co-authored RFC-0050, RFC-0054, and RFC-0056 (Nightly & Periodic CI). 50+ PRs across 5 repositories.

• RHEL CI for PyTorch — Built a downstream CI pipeline that builds and tests PyTorch nightly on RHEL 9.6 using podman containers and self-hosted runners. Includes delta-based test determination (heuristic + structural C++ call graph analysis), categorized test execution (CPU, Inductor, GPU), and CRCR integration for upstream reporting.

• PyTorch Core Contributions — Bug fixes and improvements across torch.nn, torch.distributed, and test infrastructure. Notable contributions include fixing softplus numerical instability, adding convolution overflow checks, and fixing OOT backend registration. 85+ PRs in pytorch/pytorch.

Python C++ TypeScript React/Next.js AWS Lambda DynamoDB ClickHouse Redis GitHub Actions Terraform Docker/Podman PyTorch

_{Open to collaborating on PyTorch, ML infra, and open-source CI tooling.}