Harsh Tomar

I build AI systems that actually work — from tracking tennis balls at 30 FPS to generating novel molecules

AI/ML Developer building production systems since 2024. Implemented Vision Transformers, VLMs, and LoRA from scratch in PyTorch. My Tennis Vision project earned 23+ GitHub stars for detecting players with 95% accuracy at 30 FPS. Currently building LLM reasoning systems, RAG pipelines, and agentic AI. I ship code, break things, fix them, and document it all.

Tennis Vision

Tennis Vision Demo

95% player detection • 88% ball tracking • 30 FPS

Try Demo →

DeepGuard

DeepGuard Demo

Deepfake detection • MLOps pipeline • AWS deployment

Try Demo →

Decifra

Decifra Dashboard

Fraud detection • ZenML MLOps • SHAP explainability

Try Demo →

Explore My Work

Featured Projects

7 production-ready AI systems including Tennis Vision, DeepGuard, and more

→

Research Implementations

From-scratch PyTorch implementations of ViT, VLMs, LoRA, and Reasoning LLMs

→

Additional Projects

10+ more projects covering agents, MLOps, and learning resources

→

Technical Expertise

Programming Languages

Python JavaScript Bash

ML/AI Frameworks

PyTorch TensorFlow Scikit-learn HuggingFace Transformers Keras XGBoost

Computer Vision

YOLOv5-v8 OpenCV Object Detection Image Segmentation Supervision Optical Flow PIL

Generative AI & LLM

LangChain LangGraph LlamaIndex CrewAI AG2 (AutoGen) RAG Prompt Engineering OpenAI API Gemini API Anthropic

Data Science

NumPy Pandas Matplotlib Seaborn Statistical Analysis Feature Engineering

Development Tools

Git Docker Jupyter VS Code Streamlit FastAPI Flask Gradio HuggingFace Spaces

Cloud & Deployment

AWS (EC2, S3, Lambda) GCP Firebase Netlify CI/CD

Databases

MongoDB Pinecone Chroma Vector DBs

MLOps

ZenML DVC BentoML MLflow

XAI / Explainability

SHAP LIME Grad-CAM

Professional Experience

AI Intern

i3 Digital Health May 2025 - Present

Architected an intelligent research profiling system that automatically generates researcher profiles by aggregating data from PubMed, ResearchGate, and Google Scholar APIs. Built RAG-powered search agents providing contextual recommendations and identifying potential research collaborators. Collaborated with healthcare professionals to refine search algorithms and improve researcher profiling accuracy. Working with LangChain, FastAPI, and modern AI frameworks for production deployment.

Multi-source API integration RAG-powered search Healthcare AI

Community Contributor

CNCF & Google Developer Groups Jan 2023 - Present

Active member of Cloud Native Computing Foundation participating in 15+ cloud-native technology discussions. Engaged in Google Developer Groups collaborating on machine learning initiatives, presenting at 2 tech talks on AI/ML best practices. Mentored 10+ junior developers through community workshops and open-source contributions.

15+ discussions 2 tech talks 10+ mentees

Achievements & Impact

47+
GitHub Repositories
40+
GitHub Stars
1.5+
Years Building AI/ML Systems
83%
Accuracy Improvements

Education

Bachelor of Technology in AI & Data Science

Lakshmi Narain College of Technology, Bhopal

Nov 2022 - May 2026

Relevant Coursework: Machine Learning, Computer Vision, Deep Learning, NLP, Data Structures & Algorithms, Reinforcement Learning, Statistical Analysis, Neural Networks

Recommendation

"I was impressed by Harsh's commitment and technical prowess — he attacks each challenge with enthusiasm, learning desire, and will to accomplish. His interest in Machine Learning, Computer Vision, and AI has surpassed what one might initially expect from someone at his level."
Yashvardhan Singh Software Engineer at BARCO, B.Tech. IIT Delhi
View Full Letter Internship Certificate

Open to AI/ML internships and full-time opportunities — feel free to reach out via the links above.