- Software Engineer
- InfoIMAGE, Inc.
- Mar 2024 - Present
- Brisbane, California
- Developed a new end-to-end internal Core platform using Next.js and Flask, resulting in a 80% increase in data sample lookup
- Designed Core ORM by using SQLAlchemy, to increase development time by 20% and improve security
- Utilized RESTful API to facilitate front-end to back-end communication, improving data consistency by 50%
- Integrated AG-Grid to optimize the user experience resulting 15% increase in filtering and searching process
- Automated data parsing pipeline with Python, reducing manual mapping procedures by 50%
- Assisted with implementation responsibilities by gathering job specifications, business rules, and technical deliverables through direct client interfacing
- Software Engineer
- Tricia 42 LLC
- May 2024 - Present
- Remote
- Developed a chat interface to provide helpful information from our all-in-one guide Tricia
- Lead the group of 3 developers to build front-end components for the Tricia platform using Next.js
- Implemented RAG by integrating user interactions with Supabase and VectorDB, providing a highly available and performant repository for interaction data
- By integrating Vercel AI SDK with Tricia’s backend, we’ve created a seamless chat experience that leverages the strengths of both technologies, providing users with an intuitive and efficient interface
- Provide sources of relevant forums for users to support their needs
- Machine Learning Eningeer
- Amazon Alexa Prize / University of California, Santa Cruz
- Sep 2022 - Sep 2023
- Santa Cruz, California
- Reduced training time by 23% using LoRA and lightning package to fine-tune RedPajama 3B on AWS Sagemaker
- Enhanced system speed by 34% via AWS EC2 auto-scaling deployment of RESTful server with Red Pajamas 3B
- Increased 30% user ratings by fine-tuning RedPajamas 3B with dialog history for personalized questions
- Over three months, we achieved a 10% increase in user evaluations by implementing an APL (Alexa Presentation Language) detail template with text that auto-scrolls in sync with the voice-over
- Through the design and A/B testing of multiple variations of dialog managers and APL templates, we successfully attained a 13.5% improvement in satisfactory evaluations, striving to unveil the ultimate user experience
- Data Sciencist Intern
- HPC-FAIR/Lawrence Livermore National Laboratory
- Jun 2022 - Sep 2022
- Livermore, California
- Conducted in-depth research on machine learning models with cohorts, presenting a range of innovative approaches for analyzing the SARS-CoV-2 Inhibitors dataset
- Accelerated 25% in data collection speed to extract data from diverse web sources by leveraging Scrapy’s parallel functionality
- Fine-tuned CodeBERT with DataRaceBench source code to capture parallel code style syntax to achieve the capablity to adapt parallel syntax in C/C++
- Improved 10% CodeBERT training speed by freezing the hidden layers to reduce trainable weights
