Prawin Kumar
AI Systems Engineer
Summary
AI Systems Engineer who designs and builds end-to-end AI systems for real-world problems across sectors. Experienced across machine learning, retrieval-augmented generation (RAG), LLM applications, data pipelines, model training and inference, API design, containerization, and secure production deployment. Builds adaptable AI products and infrastructure, with current public work spanning legal research, construction intelligence, scientific computing, and developer education.
Education
BSc in Data Science & Artificial Intelligence — IIT Guwahati
Research Experience
Research Intern — IIT Guwahati LinkedIn Post
May 2026 – July 2026
Worked under Dr. Rajiv K. Kar at the Jyoti & Bhupat Mehta School of Health Sciences & Technology on CASCADE, a browser-native computational chemistry platform for molecular sketching, 3D conformer generation, GNN-based property prediction, ADMET profiling, and protein-ligand interaction analysis.
Selected Works
Writ — Offline Legal RAG Research Assistant
Privacy-first legal research assistant with local LLM inference and hybrid retrieval
- Built an offline Indian legal research assistant combining a fine-tuned legal language model with an encrypted on-device knowledge base and hybrid retrieval; queries never leave the machine.
CASCADE — Computational Chemistry Workspace
Molecular research workspace with cheminformatics, ML inference, and WebGL visualization
- Created a 2D chemical sketcher, WebGL 3D conformer viewer, GNN/MLP atomic property prediction, ADME profiling, drug repurposing, and AI-assisted PyMOL workflows.
Project Neev — Construction AI Platform
Computer vision, generative floor planning, cost estimation, and 3D visualization
- Designed an AI pipeline for residential feasibility: plot scanning, genetic-algorithm floor plan optimization, ML cost prediction calibrated to CPWD DSR 2025, and interactive visualization for 28 Indian cities.
Craft with AI — AI-Native Engineering Course
42-lesson browser-first course for modern software engineering and AI systems
- Developed progressive modules covering Python, Linux, Git, FastAPI, Docker, LLM APIs, vector databases, autonomous agents, LLMOps, and CI/CD for AI-assisted development.
Vidhi — Indian Legal AI Research Assistant
Legal AI for statutory interpretation and Indian case-law reasoning
- Built a research assistant trained on Indian Supreme Court judgments, High Court orders, and Central Acts for legal document processing, information retrieval, and case-law analysis.
Technical Skills
Languages: Python, JavaScript, SQL, HTML, CSS AI/ML: PyTorch, scikit-learn, LLMs, RAG, NLP, GNNs, fine-tuning, embeddings, vector search Backend: Flask, FastAPI, REST APIs, Celery, Redis, SQLite Scientific: RDKit, ADMET, molecular modeling, 3Dmol.js DevOps: Docker, Linux, Git, CI/CD, Nginx, API security
Independent Work
VyapAI vyapai.tech
Independent AI Practice
Client-facing AI engineering across private infrastructure, custom RAG architectures, LLM fine-tuning, and multi-agent systems for legal, financial, construction, and research domains.