Martin Li

Plano, TX

• Shipped the fund platform 25+ portfolio managers and internal teams rely on for onboarding, due diligence, factsheets, and portfolio management, full-stack in React/TypeScript with Java and Spring Boot REST APIs on SQL Server.
• Co-led a Spring Boot and Jakarta migration across 35+ microservices, building the tracking system that surfaced the nine repositories with blocking merge conflicts and writing the team's incremental upgrade methodology.
• Sole-authored a full-stack compliance tool in Java, SQL, and React that turned a multi-day manual database process into self-service, removing engineering from the critical path for routine business operations.
• Built and own the Python automation that monitors 35+ services for uptime and audits CI/CD compliance across 35+ repositories, reporting status to the team daily.
• Release-engineered compliance-gated production deployments of 13+ services per release across on-prem Kubernetes and AWS EKS, plus a 22-service observability audit spanning four platforms.
• Cut deployment time up to 40% by modernizing services to Spring Boot microservices and containerizing delivery on Kubernetes and Docker with Jenkins CI/CD, sustaining 99.9% uptime across 30+ applications.
• Cut fund and database setup 70–80% by automating onboarding for 10+ funds with Liquibase-managed schema changes and tuned SQL Server queries.

Remote

• Modernized Java and SQL API services with OpenAPI/Swagger, stronger validation, CI hardening, and expanded test coverage, improving median API latency by about 30%.

New York, NY

• Automated secure 6–15 GB weekly financial file transfers with Linux, Bash, GPG, and scheduled jobs, cutting transfer errors by 25% and holding 99.9% data integrity.

Jersey City, NJ

• Built a QA data-visualization dashboard that cut outsourced reporting costs by 15% and took the tedium out of a recurring reporting workflow.

• FastAPI and LangGraph platform running adaptive technical interviews as a resumable, checkpointed state machine: 21 REST endpoints, 194 automated tests, and a CI-gated benchmark that measures the grader's own accuracy.
• Drew a hard line on where model judgment is allowed: deterministic code computes the facts and enforces the constraints, and the model handles only genuine judgment calls. Built and deployed solo on PostgreSQL and Redis with a React front end.

• Mobile speech coach giving real-time feedback on filler words, pacing, and progress, with the LLM (llama.rn) and speech-to-text (whisper.rn) both running on-device so nothing recorded leaves the phone. TypeScript and Expo, built and shipped solo.

• Local-first React Native and Expo app built on one idea: the hard part is starting a task, not finishing it, so nothing in the app punishes a miss. Research-grounded behavioral UX, WCAG AAA accessibility (7:1 contrast, large targets, text-to-speech, haptics, undo), and a 155-assertion suite that guards the design rules alongside the logic.

• Graduate study in Python and scikit-learn comparing backpropagation, genetic algorithms, and MIMIC for heart-disease classification, reaching about 0.945 accuracy and mapping where each method stops improving.

Taken part time while working full time. Coursework: Machine Learning; Artificial Intelligence; ML for Trading; Data & Visual Analytics; AI, Ethics & Society

• 1st Place, Cornell University Data Science Hackathon
• 1st Place, Heroes vs Hackers Hackathon
• Best Website Hack, Columbia Health Hacks