Tushar U

I’m an aspiring AI and Data professional with a strong foundation in modern data systems, Machine Learning, and Generative AI. I enjoy turning messy inputs into reliable data workflows and intelligent outputs—especially when the solution can reason, retrieve context, and drive better decisions.

In recent internships, I built an agentic AI-powered RAG system using a local LLM to retrieve and summarize sustainability policies, focusing on retrieval-augmented workflows and prompt engineering for better relevance. I also contributed to an AI-driven multimedia pipeline with timestamp-synchronized speech-to-text and LLM-based abstractive summarization.

Alongside this, I’ve supported data analytics on multi-source datasets by performing data cleaning, validation, and exploratory data analysis to ensure quality and trustworthiness for decision-making. My projects reflect the same mindset—production-style pipeline reliability, reproducibility, and ML-ready feature engineering across Bronze–Silver–Gold layers.

I’m especially motivated by building practical NLP and computer vision solutions, from transformer-based intent automation and confidence-driven human-in-the-loop escalation to a hybrid YOLOv11 + U-Net approach for mitosis detection. I bring a disciplined, learning-first ethos—backed by consistent achievements (I Rank in Vidwat Bharatanatyam, 2022) and experience in multiple undergraduate technical hackathons.