-Built a Remix popup for user feedback and analyzed responses via PostHog.-Integrated a RAG-based pipeline, improving accuracy across 80 questions.-Assisted in transitioning the core LLM from GPT-4 Turbo to GPT-4 by comparing over 900 Q&A pairs.-Fixed bugs in the Django backend, boosting product accuracy.-QA’d 250+ LLM calls to ensure high-quality performance.

- Scraped the Indian Govt Patents Website and integrated Retrieval Augmented Generation, and implemented metadata filtering, dense vector and sparse search (BM25) on a database with more than 400,000 documents- Working on using Vector Embeddings generated by Sentence Transformers and storing in Weaviate and QDrant- Made a OpenAI proxy server to monitor calls being made to OpenAI for better tracking and setting limits- Working with Langchain Agents and Tools to provide solutions to integrate AI into the workflow of client websites- Made WhatsApp and Instagram bots along with Text to Speech, Speech to Text and Voice Cloning methods- Worked with GCP and Azure for production level deployment along with PostgreSQL and Redis with a Flask backend