Open to work

My open-source journey

I started by fixing the tools that broke under my own projects. Now I read other people’s codebases on purpose: on-device runtimes, model converters, LLM gateways and the Swift standard libraries, from Google, Microsoft, Apple, NVIDIA and Hugging Face. Every pull request below is pulled live from GitHub, across both of my accounts, sandeepvijayarao09 and mlzerotoheroyoutube.

— PRs merged by maintainers
— PRs in review upstream
— upstream repos contributed to
— combined stars of those repos
— commits merged into those repos

Where the work lands

On-device runtimes

LiteRT · ONNX Runtime · MLX Swift · TFLite Micro

Kernel and backend bugs: int16 comparison kernels in LiteRT, a Core ML model-cache leak in ONNX Runtime, a zero-size-array crash in MLX Swift.

Model tooling

ONNX Runtime GenAI · LiteRT Torch · MLX Swift LM · Argmax

Converters and loaders: Gemma 3 RoPE scaling in the model builder, a silent no-op in the NPU compiler, GLM-OCR image handling, M5 and A19 device support.

LLM and agent infrastructure

LiteLLM · NVIDIA auto-ontology · Semantic Router

Error mapping for Anthropic streams in LiteLLM, keeping PII sample values out of text-to-SQL prompts, and encoder defaults in Semantic Router.

Swift and SDKs

Swift Collections · RunAnywhere · Diffusers

A B-tree invariant in Apple’s SortedCollections, cross-platform SDK fixes and ABI documentation in RunAnywhere, and 22 dead links in the Diffusers docs.

How it started

  1. Early 2026
    Building Anu on LiteRT

    Shipping Gemma 4 on an iPhone for the Apple Swift Challenge meant living inside LiteRT, MLX and Core ML, and finding where they break.

  2. Mar 2026
    First pull request

    A memoized solver for a classmate’s CodeCorrect project. Small, but it was the first time my code went through someone else’s review.

  3. Nov 2026
    First upstream merges

    RunAnywhere and Hugging Face Diffusers merged my first fixes within a day of each other.

  4. Now
    Going wider

    Fixes in review at Google, Microsoft, Apple, NVIDIA, ml-explore and BerriAI. Each one starts with a reproducible failing test.

Every pull request

On GitHub
  1. Loading from GitHub…

Open to work

Looking for a team that ships AI.

AI engineering and software internships and new-grad roles. MS CS at Northeastern, graduating 2027, based in San Jose.

Get in touch Resume