Hi I am Ankur Singh,
Currently, I’m working as an AI Solutions Engineer at Intel. My part of my job, I had to migrate some CUDA samples to Intel’s SYCL. This experience has ignited my passion for GPU programming. It has led me to explore tools like Triton and torch.compile, which make CUDA programming more accessible for Python developers, and has also sparked my interest in C++. Checkout my notes in Today I Learned (TIL) tab.
I earned my Master’s in Software Engineering from San Jose State University in May 2024. During my studies, I served as a research assistant under Dr. Wu and Dr. Liu at the SJSU Research Foundation. My research focused on Multi-Modal modeling, Federated Learning, Traffic Flow Prediction, and real-time edge deployment, providing me with hands-on experience in these cutting-edge areas.
With 4 years of prior professional experience in AI/ML, Software Engineering, DevOps, Cloud, and Databases, I embarked on my journey by founding AI Adventures, a startup providing AI/ML solutions. Later, as an ML Team Lead at Zoop.One, I orchestrated the successful launch of four ML services, housing 20+ deep learning models, handling over 2M+ requests/month. All this in just ten months. My approach, grounded in first principles, emphasized improving developer experience through process building, establishing best practices, and setting up efficient ML infrastructure, leading to rapid prototyping and significantly reduced Time To Market (TTM). I’m really proud of my work at Zoop.
Beyond my professional endeavors, I am an avid reader and actively participate in competitions, winning multiple Kaggle competitions and hackathons. I’ve contributed to several open-source projects and created two Python packages: colab-everything and torchserve-python, aiming to give back to the community. I aspire to give a talk at a Python Conference someday.
Data Science
85%
Pandas, SQL, BigQuery, PySpark, Sklearn
DL Stack
90%
Pytorch, Triton, Transformers, PeFT, Accelerate, CUDA
LLM Stack
85%
DeepSpeed, PyTorch Distributed, vLLM, LlamaCPP
MLOps
85%
MLflow, ONNX, OpenVINO, TorchServe, TensorRT
Programming
90%
OOPs, Design Patterns, Packaging, Unit Testing
CI/CD & DevOps
85%
Git, GitHub Actions, Docker, K8s, AWS, Terraform
Databases
80%
Postgres, SQLite3, MongoDB, Redis, ElasticSearch
API Development & Web-apps
85%
FastAPI, Flask, Firebase, Appwrite, Steamlit, Gradio
Develop, benchmark, profile, and optimize various AI workloads on Intel hardware using Intel's optimization stack. This includes workloads such as distributed training, QLoRA, quantization, custom PyTorch kernels, and LLM deployment. Additionally, responsible for migrating CUDA samples to SYCL and helping maintain over 30 AI code samples.
Optimized the YOLOv8 model for object detection and segmentation to enable real-time inference on NVIDIA Jetson devices connected through ROS. Contributed to developing innovative strategies for Knowledge Distillation and Federated Learning, with a focus on regression problems. Additionally, worked on multi-modal modeling, including image and audio tokenization to prepare these modalities for input to LLMs, and using LLMs to generate image and audio during inference.
Led the strategic development of machine learning initiatives by launching four core ML services, which include over 20 deep learning models and handle more than 2 million requests per month—all within just ten months. Established best practices for MLOps, developed micro-frameworks, and implemented automation processes, with emphasis on agility and an exceptional developer experience.
Led client project development, overseeing the entire lifecycle and collaborating closely with clients. Created five comprehensive courses covering Python, Data Science, Machine Learning, Computer Vision, and Deep Learning. Established AI clubs in several colleges across the city, fostering a vibrant community through workshops and interactive sessions.