Developed and prototyped new Gemini powered features for Gmail:- RAG pipelines, LLM Chain of Thought / Planning / Tool Calling, Synthetic Data Generation- C++ Infrastructure

Leveraging multimodal LLMs for auto evaluation tasks

• Developed synthetic user data generators for Siri evaluation and testing• Added new features to and resolved bugs on Swift native infrastructure for Siri testing, including integration of Calendar, Contacts, and iMessage into experiment contexts• Created custom native AWS S3 data uploader with Swift, enabling large scale uploads of Siri experiment results into cloud• Created Python pipelines with Streamlit UI that improved efficiency of multiple internal client teams

Developed the knowledge base for SVL’s robotics and AI problem solving project. Singlehandedly created the website and server to display this database. • Sole full stack developer of frontend and backend of knowledge base website• Utilized Python Flask for backend• HTML, JavaScript, CSS, jQuery for frontend• Data analysis and manipulation of .json files and spreadsheets with Python• Mentored by SVL’s postdoc researchers with regards to natural language processing/generation

• Analyzed real data from sensors attached to cups and shoes, and built a neural network model to differentiate the motions of one being drunk from random motions. Achieved >95% accuracy.• Utilized LSTM recurrent neural network and other algorithms to build models of human walking. • Developed web frontend and backend for real-time demo for investor showcases. • Competitive analysis on 50+ healthcare AI and wearable companies. Contributed to DawnLight’s product differentiations and marketing strategy.