Professional experience, open-source programs, visiting faculty engagements, individual contract work, and community recognition.
Working on self-improvement pipelines for AI Agents, iterative learning workflows, evaluation-driven optimization, and autonomous improvement strategies. Published Evret, a lightweight Retriever Evaluation Framework. Developed on-device AI search orchestration and self-hosted LLM inference for offline forensic and defense workflows.
Deployed AI models on an AI Marketplace and developed LLMs and AI Agents for enterprise PoCs across education, banking, finance, and CPG. Led the AI engineering team, built Panda Coder 13B, led Buddhi-128K Chat Model, and supported community programs across a 300K+ global community.
Worked under Dr. Cyril Prasanna on applied research projects including hyperspectral image compression on FPGA using MATLAB and object tracking with Deep SORT and TensorFlow.
Developed a multilingual news dataset with 64,000+ QA pairs across 6 languages for fine-tuning LLMs using Self-Instruct. Optimized fine-tuning and inference with PEFT and Flash-Attention 2, and implemented a RAG inference pipeline with guardrails and evaluation.
Built multi-channel medical image support for caMicroscope, enabling researchers to map arbitrary image channels to RGB for visualization. Worked with TIF and hyperspectral images, Flask APIs, OpenSeadragon/DZI tile visualization, Docker Compose integration, and JavaScript changes inside the caMicroscope codebase.
Collaborated with leading AI framework, vector database, and LLM observability ecosystem partners:
Delivered hands-on lectures, interactive sessions, and specialized workshops on NLP Search Algorithms and optimizing Machine Learning algorithms.
Conducted expert teaching sessions, guest lectures, and technical deep-dives on PyTorch and Deep Learning.
Delivered 5+ FDP talks.