Minh Trinh

Building an end-to-end ML-powered system for automatic root cause analysis for failures in Tesla Gigafactory.

- Research large language models (LLMs) for synthetic data generation with NLP techniques: fine-tuning, tokenization, prompt engineering - Developing evaluation pipelines for generative models, such as GANs, quantifying synthetic data’s privacy and utility

Publication, first author: https://www.frontiersin.org/articles/10.3389/frdem.2024.1332928/

- Developed graph neural networks for data processing, for use in Next Generation Interceptor/Golden Dome-related contracts - Sped up classic algorithms by orders of magnitude using efficient JAX-based machine learning code

- Developing an intelligent video dataset curation product using state-of-the-art data pipelining tools, cloud technology, and AI models

- Developing data and machine learning pipelines for generative AI models in film using cloud compute and orchestration software - Work in professional software development setting: sprints, standups, code review, cloud tools, CI/CD workflows, etc.

- Led a team of 10 engineers to create a machine learning-based pipeline to generate a dataset of foreign language videos - Used computer vision models for face detection, segmentation, classification, audio models for speech analysis and generation

- Developed multiple Python applications for testing 3000+ radios, released into production after a thorough code review process - Composed official software specification documents, used as guidelines for team’s development and QA processes - Contributed to C++ codebase for embedded microcontrollers for electronic products - Developed software programs for data manipulation and visualization in Python and SQL; released into production

- Published paper as first author in Harvard's Journal of Emerging Investigators: https://emerginginvestigators.org/articles/22-171 - Proposed efficient CNN architectures to diagnose plant diseases via leaf photos and proved they outperform conventional CNNs - Developed mobile app with React Native to diagnose plant diseases in real time; won 3rd among 50 in Congressional App Challenge

Advanced to California State Science Fair; won 3rd in Orange County Science Fair Developed deep neural networks with Q-learning to identify optimal stock trading strategies, including sentiment analysis

Junior Standing

Linear Algebra, Ordinary Differential Equations