Arihant Sheth

- Improved duplicate bill detection recall by 30% and cut false positives by 83% through optimized retrieval - Drove adoption of structured output schemas as a BILL-wide best practice, accelerating AI development - Reduced targeted deployment costs by ~90%, enabling scalable rollout without sacrificing quality

- Adapted Optimization by PROmpting (arxiv:2309.03409) from Google DeepMind for math modeling - Formulated an adaptive & data-driven starting-point heuristic that reduced optimization time and cost by 50% - Integrated RAG with chemical & power domain-specific knowledge to enhance mathematical reasoning accuracy

- Automated product lifecycle by Product Requirement Document (PRD) generation and competitive analysis - Created a novel token management technique that halved PRD generation cost - Engineered a seamless RAG pipeline to allow the LLM to request information beyond its internal knowledge

- Spearheaded the implementation of a cloud computing solution using Google Cloud Platform for a sports training company, enabling efficient storage of sport-specific videos in buckets. - Integrated sports videos seamlessly into a Flutter app, collaborating closely with the frontend team to ensure smooth functionality and user experience. - Developed an intuitive and user-friendly admin dashboard to monitor coaches and track video viewership, along with assigning new videos.

- Established and nurtured a dynamic data science community comprising 250+ students through strategic organization of hackathons, workshops, and seminars featuring leading industry experts. - Personally mentored 100+ aspiring engineers through workshops, lectures, and more, empowering them to unleash their capabilities and cultivate a profound interest in data science.

- Implemented a Docker-based runtime environment for the research paper "Text2Light," enabling the generation of 3D world environments from textual user input. - Collaborated with a team to explore the potential application of the Text2Light model in creating immersive Virtual Reality (VR) experiences.

- Deployed an end-to-end ETL pipeline to automate the stock trading process using Apache Airflow for Python - Developed a backtesting algorithm that autonomously selects the best hyperparameters each week

- Worked in the technical team for the year, and built an IoT-powered LED cube that served as a technical attraction for the event.

Summer '24: 11-637 Foundations of Computational Data Science (A) Fall '24: 10-601 Introduction to Machine Learning (A) 11-785 Introduction to Deep Learning (PhD) (A+) 05-849 Interactive Data Science (A) 11-631 Data Science Seminar (A) Spring '24: 15-619 Cloud Computing ☠️ (A+ 🙂) 11-695 AI Engineering (A+) 11-681 AI Venture Studio (Founded a startup and built traction!) (A) Capstone: Fine-tuning LLMs for secure code generation (A) Fall '25: 11-667 LLM Methods and Applications (A) 10-714 Deep Learning Systems (A+) Capstone: Fine-tuning LLMs for secure code generation (A+ 🙂) Intro to Spring Board Diving!