Designed, developed, and deployed 3 GenAI MVPs, including Planning AI agents, Comment Insights systems, and Retrieval-Augmented Generation (RAG) agents.

1. Built and led team of 6 AI/ML engineers, delivering AI services and solutions for 5+ enterprise clients across medical, e-commerce, ed-tech, and real estate domains.2. Implemented OCR-based document processing using PaddleOCR, PyMuPDF, YOLO, and OpenCV, enabling layout analysis and classification of 140+ medical record categories with custom domain rules.3. Engineered semi-automated pipelines for medical event chronology and summarization using LangChain and Llama 2/3, reducing… Show more 1. Built and led team of 6 AI/ML engineers, delivering AI services and solutions for 5+ enterprise clients across medical, e-commerce, ed-tech, and real estate domains.2. Implemented OCR-based document processing using PaddleOCR, PyMuPDF, YOLO, and OpenCV, enabling layout analysis and classification of 140+ medical record categories with custom domain rules.3. Engineered semi-automated pipelines for medical event chronology and summarization using LangChain and Llama 2/3, reducing manual processing time from days to hours while protecting data privacy with on-premises GPUs.4. Developed e-commerce product-matching systems leveraging OpenAI, LangChain, and statistical analysis to standardize and normalize product features across unstructured data and diverse units of measurement. Show less

1. Led dailyhunt AI Labs of 10+ members leading efforts in recommendation, NLP etc.2. Implement and deploy recommendation framework, serving personalised content to 300M users across 13 Indian3. languages including 7 candidate generators models (content-based, collaborative filtering, graph based etc.), DL based ranking model working together in a single pipeline.4. Built NLP-driven solutions for news content in 13 indian languages, including classification, keyword detection… Show more 1. Led dailyhunt AI Labs of 10+ members leading efforts in recommendation, NLP etc.2. Implement and deploy recommendation framework, serving personalised content to 300M users across 13 Indian3. languages including 7 candidate generators models (content-based, collaborative filtering, graph based etc.), DL based ranking model working together in a single pipeline.4. Built NLP-driven solutions for news content in 13 indian languages, including classification, keyword detection, graphbased story grouping, and entity linking with knowledge graphs for precise content recommendations.5. Developed large-scale feature processing pipelines using Spark, TensorFlow, Airflow, and GCP, processing 1TB+ data per hour to continuously refine recommendation models for Dailyhunt and Josh.6. Trained and deployed a YOLO4-based NSFW detection model to moderate 10k+ short videos daily on Josh, outperforming benchmarks on indian datasets.7. Created a cold-start recommendation system for new users to quickly identify user interests within five interactions to improve retention. Show less