Kirtan Shah Email & Phone Number

United States

San Mateo, California, United States

• Boosted grooming classifier precision 16% and F1 score 4.1% by engineering prompt optimization pipeline for LLM classifiers using Bayesian search and ChainOfThought with DSPy.
• Processed 140,000+ lines of real-world multilingual chat data using Python and Pandas, stratifying datasets across multiple languages and labels to ensure robust model training.
• Integrated with knowledge distillation pipeline, enabling enhancements for all safety classifiers.
• Collaborated with ML Platform team to cut experiment iteration time 3x with Kubeflow, Ray, vLLM.

San Diego, California, United States

• Developed user and machine management system using MySQL, React.js, Flask/gunicorn (Python).
• Assisted students with CAD, 3D printing, laser cutters, woodshop tools and electronics equipment.

Seattle, Washington, United States

• Spearheaded automation of global expansion of Kinesis Data Streams for DynamoDB (KDSD), streamlining build and deployment processes across diverse AWS regions.
• Leveraged Ruby, Java, TypeScript to automate infrastructure as code, reducing developer time 75%.
• Swiftly integrated into KDSD team, mastering new tools and adapting to dynamic AWS ecosystem.
• Collaborated cross-functionally with software and system teams to ensure project delivery.

Seattle, Washington, United States

• Enabled savings of $17 million and reduced storage footprint (100PB+) of logs over 7 years by implementing log filtering system for DynamoDB Global Tables.
• Updated high throughput logging library, enabling all DynamoDB teams to reduce storage footprint.
• Utilized Junit to rigorously test Java code changes, especially in high-impact internal packages.
• Collaborated with senior and principal engineers to develop and iterate on designs.

Palo Alto, California, United States

• Boosted live-chat team productivity by collaborating with product development and implementing quality of life features like automatic navigation to client’s product page.
• Developed new features in order flow using Vue.js/Ruby/HAML for UI elements, TypeScript for business logic, and Ruby on Rails for RESTful API endpoints.
• Ported analytics code to internal codebase, improving fidelity of reporting from 95% to 100%.
• Automated sync of localization resources via bash/git hooks to improve developer efficiency.
• Quickly adapted to closed source development environment, integrating features into main product in week 1.

Saratoga, California, United States

• Converted financial assets spreadsheet into SQL database using custom Python scripts tostore data robustly and scalably.
• Gathered feedback from city workers / financial lead to customize an intuitive React.js andElectron.js front end

Master Of Science - Ms, Machine Learning / Data Science

Bachelor'S Degree, Computer Engineering, 3.9/4.0

Activities and Societies: Engineers For Exploration (E4E), Institute of Electrical and Electronics Engineers (IEEE), Association for.

High School Diploma