Tejaswi Pratap

I’m an AI/ML Engineer with 1.5+ years of experience building production LLM systems, behavioral analytics copilots, and Retrieval-Augmented Generation (RAG) pipelines. I focus on turning real user and product signals into grounded, reliable AI experiences.

At .ai, I designed and built Overlord from scratch, adding a real-time event-capture layer between the frontend and the LLM backend to observe page navigation, feature usage, dwell time, and interaction patterns across SaaS report pages. I also engineered a session-scoped dynamic context retrieval pipeline with session IDs and continuous context refreshing so my assistant stays grounded to the user’s current view in real time.

I built CreatorGPT, a RAG-based Q&A system for influencer-marketing domain questions, including a YouTube transcript scraping and ingestion pipeline and an LLM-driven synthetic Q&A dataset to serve as the knowledge base. I improved retrieval precision using question-to-question semantic matching—embedding and matching against a question bank to retrieve the top 5–8 semantically closest questions and synthesize high-confidence answers.

I’ve contributed to agentic AI workflows, including role-specific prompts and logic for an Outreach & Brief Agent and a Negotiation Agent using the Gemini API. I also integrated Supabase for outreach tracking and led Instagram hashtag performance analytics with automated scraping pipelines using Selenium and Playwright in Python.