Ankur Singh
Sr System Software Engineer at NVIDIA
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAnkur Singh is a senior system software engineer with a decade of experience building and optimizing AI/ML systems, currently focused on LLM post-training and inference at NVIDIA after impactful AI solutions work at Intel. He has a strong track record delivering production-grade GenAI workflows—QLoRA, INT4/8 quantization, and CPU/XPU inference optimizations—that produced up to 3× performance gains through targeted profiling and PyTorch-level tuning. Ankur has architected modular, microservices-based LLM applications (RAG agents, Q&A chatbots, video search) deployed with Docker/Kubernetes and a rich stack including vLLM, LangChain, ChromaDB, and Triton. He led ML teams and launched multiple ML services at Zoop.one, founded AI Adventures to train and scale AI talent, and contributes to open source work on projects like torchtune, improving model evaluation and usability. Based in San Jose with an M.S. in Software Engineering from SJSU, he blends hands-on optimization expertise with product-minded engineering and a habit of turning research techniques into deployable tooling. A less obvious strength: he repeatedly bridges edge-to-cloud inference workflows, from Jetson/edge deployments to distributed CPU fine-tuning, making him effective across the full ML lifecycle.
10 years of coding experience
7 years of employment as a software developer
Bachelor of Engineering - BE, Information Technology, 7.76 CGPA, Bachelor of Engineering - BE, Information Technology, 7.76 CGPA at College of Engineering PuneSan José State UniversityHSC, Science, 94.7%, HSC, Science, 94.7% at Sri Chaitanya Junior College
Hindi, English, Marathi
Github Skills (9)
model-building10
configuration-management10
pytorch10
machine-learning10
tokenizer10
tokenize10
python9
documentation8
evaluation8
Programming languages (8)
Github contributions (5)
meta-pytorch/torchtune
Jan 2025 - Apr 2025
PyTorch native post-training library
Role in this project:
ML Engineer
ML EngineerContributions:21 reviews, 13 PRs, 72 comments in 3 months
Contributions summary:Ankur primarily contributed to the development of the torchtune library by implementing and integrating various features related to model evaluation, dataset handling, and model building. They added configurations for evaluating the QWEN2_5 model and refactored modules and tokenizers. The user also focused on improving the library's usability by incorporating logging configurations, implementing dropout layer disabling, and updating documentation. Their work demonstrates a strong understanding of model training, tokenization, and configuration management within the PyTorch ecosystem.
post-trainingpytorch
Ankur-singh/colab_everything
Oct 2020 - Oct 2022
Python library to run streamlit, flask, fastapi, etc on google colab.
Contributions:19 commits, 5 PRs, 18 pushes in 1 year 11 months
fastapiflaskgoogle-colabpythonstreamlit
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.