i am an engineer interested in the intersection of machine learning, systems, and performance. more specifically i'm interested in compiler optimizations, distributed systems, distributed training, and machine learning infrastructure. currently, i am working on pytorch at meta ai where im building pytorch 2.0 compiler's dynamo and inductor. i'm also a junior studying computer science student at georgia tech. i'm also incoming @ google where i'll be working on serving ML models for youtube's distributed reccomendation systems + ranking models. previously i worked on many different things such as building a webapp karuna at mit pathcheck which got 23.4 million users at its peak, was a machine learning researcher working on computer Vision / ai inference under Dr. Kyle Keane PhD, did swe internships at amazon, hubspot, and union pacific railroad.
i care about building fast, scalable systems for foundation models — everything from training infrastructure and parallelism to low-latency inference at production scale.
more recently, I’ve been contributing to open-source compiler infrastructure in PyTorch 2.0, focusing on graph capture and low-level optimizations in TorchDynamo. My work involves improving code generation and execution paths to accelerate model performance through better tracing, graph transformations, and runtime-aware specialization across the Dynamo-Inductor stack.
i'm also interested in early to mid stage startups and quant finance.
outside of cs, i watch tons of movies letterboxd and watch basketball with my friends
Languages: Python, C++, Go, TypeScript, React, Java, JavaScript, C
Technologies: PyTorch, TensorFlow, NumPy, Triton, Hugging Face, AWS, Docker, Redis, Kubernetes, PostgreSQL
some projects i've built in the past:
coming soon!
feel free to reach out at sidharth.subbarao17@gmail.com!