Headline

ML Engineer | Computer Vision & Medical Imaging · LLM & RAG Systems | AI Research @ IIT Ropar | PyTorch · LangChain · HuggingFace | Kaggle Master

About

I build annotation-efficient AI that ships in clinical and production environments — not just benchmarks. At IIT Ropar, I designed a semi-supervised cardiac MRI segmentation framework that cut boundary error by 50% (HD95: 10.62 → 4.85) using only 20% labeled data, outperforming fully-supervised baselines. The result was a 1.8M-parameter model deployable at the edge. On the LLM side, I built a production RAG pipeline indexing 7,700+ ArXiv papers at sub-2s latency with a 94% faithfulness score and under 6% hallucination rate. My edge: I operate at the intersection of scarce-label learning and efficient inference — a rare combination in applied medical AI. Stack: PyTorch · LangChain · HuggingFace · OpenCV · FastAPI · Docker Published: 2× Springer (skin lesion analysis & augmentation) Looking for ML Engineer or Computer Vision Engineer roles in healthtech or applied AI — preferably teams where research meets production. Let's talk: aniketkolte0406@gmail.com

Location

India

Work experience

Research Intern at Indian Institute of Technology, Ropar, Jun 2024 - Feb 2025 · 9 mos, Rupnagar, • Developed an uncertainty-guided semi-supervised framework, reducing cardiac MRI boundary error by over 50%. • Engineered a comprehensive PyTorch pipeline, cutting model development cycles by 40% for efficient deployment. • Validated the effectiveness of annotation-efficient deep learning, demonstrating superior performance on limited labeled data.

Research Intern at Vishwakarma Institute of Information Technology, May 2023 - Aug 2023 · 4 mos, Pune, Maharashtra, India, • Redesigned a SegNet encoder-decoder CNN, improving skin lesion segmentation by +7.38 pp Jaccard and +5.30 pp Dice. • Optimized architecture depth for edge deployment, enabling resource-efficient clinical inference validated across dermoscopic images. • Contributed to two peer-reviewed publications on AI-based augmentation techniques for dermatology datasets at Vishwakarma Institute of Information Technology.

Education

Bachelor of Technology - BTech, Artificial Intelligence and Data Science, Vishwakarma Institute of information technology