Harshit Kumar

Machine Learning Engineer
Open to Work (contact me on LinkedIn)

Machine Learning Engineer with over 3 years of experience specializing in optimizing model performance and tuning low-precision neural networks. Successfully improved model accuracy training speed, and accelearted data science efficiency by implementing advanced ML techniques and leveraging distributed training systems. Experienced in end-to-end ML pipeline optimization and real-time deployment for surveillance applications. Have worked with Explainable AI techniques and Exploratory Data Science in genomic research. Worked in Computer Vision field for 2 years in Intelligent Video Analytics for Smart Cities. Have done several projects in NLP, and utilizing Large Language Models (LLM) based solutions for Generative AI tasks.

Interests 👨🏻‍💻: Machine Learning, Deep Learning, Computer Vision, Natural Language Processing, Data Science, MLOps, Generative AI, LLM
Hobbies 🔭: playing guitar, piano, stargazing, traveling, sudoku, poetry.

Harshit's Github chart — GitHub contributions

Experience

Deep Learning Research Co-op - The Jackson Laboratory

Boston, MA, US, Jul 2023 – Dec 2023

Researched and applied Explainable AI methods for Graph Neural Networks.

Technologies used: PyTorch Geometric, PyTorch Captum

Machine Learning Engineer - Vehant Technologies

Delhi, India, Jun 2019 – Aug 2022

Intelligent Video Analytics for Smart City: computer vision, object detection, tracking, reidentification, license plate recognition (ANPR OCR), action recognition, pose estimation.

MLOps: Distributed Training, Deployment, ETL, and Optimization of multi-GPU end-to-end ML pipelines (PyTorch, TensorFlow, Nvidia TensorRT, CUDA, DeepStream, MLFlow, ONNX).

Accelerated Data Science: Low precision, Quantization, Pruning, Nvidia RAPIDS.

Mentoring: 2 fellow teammates Edge AI topics.

Computer Vision R&D Intern (continued)

Jun 2019 - Jul 2020

Researched on deep-learning-based monocular depth estimation methods for under-vehicle object detection.

Technologies used: PyTorch, TensorFlow, NVIDIA DeepStream, OpenCV, MLOps