Experiences

• Leading a new initiative on ML-generated avatars and text-to-avatar generation.
• Co-authored peer-reviewed research papers accepted to workshops at top venues (ICSE, CVPR).

• Researched the significance, quantitative impact, and market awareness of foreign economic events on US commodities.
• Designed and implemented a high-speed, interactive full-stack tool that helps traders visualize counterparty toxicity, trading strategies, and positions by analyzing past counterparty actions, using SQL, caching, and Plotly.

• Developed a new trading strategy in the US equities market by researching international market signals.
• Enhanced forecasts of options volatility and built a model that surpassed implied market predictions by analyzing earlier-in-the-day predictors.

• Built a full-stack project on Route 53, CloudFront, API Gateway, S3, Lambda, and DynamoDB.
• Created a tool that decreased incoming support tickets by 90% and support response wait times by 95%.
• Wrote React + TypeScript front-end, Java AWS Lambdas, and CDKs.

Built an AI neural-network NLP translator used by 5,000+ people.

• Designed and shipped a web crawler that visited 6M+ relevant websites and a high-quality Interlingua translator using Python, BeautifulSoup, PyTorch, and CUDA.
• Created the Interlingua Corpus Project, with over 1.2M quality-controlled Interlingua sentences (a 4000%+ increase over the previous largest corpus) plus 80,000+ parallel Interlingua-English sentences.
• Built and optimized the first public, open-source online Interlingua–English translator, scoring a 42.54 BLEU score using techniques like normalization and proper-noun replacement.
• Learned to develop with React; manage Git, Jupyter, and remote servers; and deploy Python on Heroku.

• Researched safety, security, and privacy in reinforcement and deep learning with Prof. Dawn Song, including adversarial robustness, anomaly detection, and data augmentation in Python and PyTorch.
• Co-created an educational course on machine-learning safety.

• Built MATLAB and Python programs to analyze and visualize tumor growth and motion in mice over time.
• Created an app that reduced daily post-scan analysis time from 1 hour to 2–3 minutes.
• Generated 50+ visual 3D models in MATLAB highlighting tumor locations inside mouse bodies.