Jihui Tan

AI Software Engineer

Cleolabs AI (startup)•  March 2024 - August 2024•  Chicago, IL

Ideated and built web app in Python and JavaScript for children with autism to engage with AI speech therapist, leveraging Node.js, Quart, Azure, MongoDB, SoulMachine, DALL-E, ChatGPT, and Llama. Collaborated with speech therapists to devise metrics for tracking children's communication improvements on platform.

Website Development Assistant

University of Chicago, IT Services•  September 2023 - May 2024•  Chicago, IL

Developed and maintained over 100 departmental WordPress-powered websites throughout university within Web Services team, resolving 117 support tickets and offering consultation sessions to 58 academic directors and professors.

Software Engineer - AI Consultant

Outlier AI•  February 2024 - May 2024•  Oakland, CA

Designed over 300 prompts to break large language models, assessed model responses through analytical commentary, and wrote over 15,000 lines of code in Python(Numpy, Pandas), SQL, and JavaScript based on constructed prompts.

Data Science Intern

Chicago Trading Company - Data Science Clinic, University of Chicago•  March 2024 - May 2024•  Chicago, IL

Led development of automated X data collection and preprocessing pipeline in Python, and created API endpoint for team data access to expedite model development, leveraging Flask, AWS EC2, S3, and Rettiwt API. Increased data volume tenfold and model accuracy by 3%.

Software Engineering Intern

GridLink•  January 2024 - March 2024•  Chicago, IL

Designed and built Python Flask-based benchmarking system to assess performance of EV fleet charging algorithms, facilitating automated test generation, time and memory data measurement, storage, and visualization in matplotlib.

Data Science Intern

Morningstar, Inc - Data Science Clinic, University of Chicago•  September 2023 - December 2023•  Chicago, IL

Conceived and developed an ensemble-based fact-checking model in Python within four-member team using ChatGPT API to verify AI-generated reports from analysts' notes, achieving 80% precision and 95% recall in error identification.