Lily Faris
AI + Sustainability · Full-stack Problem Solving · Growth-OrientedRecent CS Graduate | Part-time AI Development Consultant | Seeking Opportunities
Recent CS Graduate | Part-time AI Development Consultant | Seeking Opportunities
I am a computer scientist working part-time as a contracted AI development specialist with Outlier AI. I recently completed my M.S. in Computer Science at UC Santa Cruz, where my thesis work focused on AI applications in sustainability (RootWise).
Prior to grad school, I earned my B.S. in Computer Science at UCSC where my thesis involved exploring AI-integrated market simulations and applied vision transformers to score audio spectrograms in an interdisciplinary research context.
My academic and research interests include responsible AI, sustainable computing, and the development of intelligent systems that address real-world social and environmental challenges.
You can view my resume [here] and GitHub [here]. I'm actively seeking full-time roles in software, AI/ML, or applied research.
An AI-powered chatbot that promotes sustainable, zero-waste cooking by combining NVIDIA’s embedding models, OpenAI’s LLMs, and the Institute for Functional Medicine (IFM) toolkit. Built with FAISS, custom RAG pipelines, and contextual personalization, RootWise achieved 82.5% success in personalization and 100% dietary compliance. Evaluated across Self-BLEU, relevance, and groundedness metrics.
Awarded in the NVIDIA × LlamaIndex Developer Contest.
Built an AI-driven market simulation for my undergrad thesis at UCSC's LEEPS Lab, explored how widespread access to AI-driven trading strategies could impact the structure and behavior of financial markets. I built a configurable continuous double auction (CDA) simulation supporting multiple trader clients, and developed a natural language strategy interpreter using OpenAI’s ChatGPT API and LlamaIndex (RAG). The system enabled users to deploy trading strategies via plain-text instructions, simulating the reduced cost of trader acquisition in an AI-accessible market.
githubEngineered a machine learning web app that transcribes .wav audio files into precise sheet music (.pdf) using computer vision. Trained on 12,796 spectrograms from the NSynth dataset with a vision transformer architecture (Dosovitskiy et al.), achieving 90% training accuracy. The project surfaced key challenges in generalization, with 4% testing accuracy—highlighting future opportunities in model tuning and dataset refinement.
githubIndependently developed an educational game highlighting the ecological role of native pollinators in sustaining environmental health. Designed to inform players on ecosystem stewardship and prototype an interactive digital approach. Collaborated with the Ann Arbor Environmental Commission (as a Youth Commissioner) to discuss integrating projects like this into education campaigns on sustainability.
download the html