Software Engineer · Backend Systems, ML Infrastructure & Applied AI
Computer Science & Statistics at Harvard University · Class of 2027

I build production systems for data-intensive and operationally critical workflows. Across security, machine learning, payments, and education, I have owned software from architecture and implementation through deployment and optimization, with a focus on reliability, performance, and measurable impact.

Cambridge, MA

Languages

Backend, APIs & Databases

Machine Learning & AI

Frontend, Cloud & Tooling

• Engineered a real-time collaborative IDE using Yjs, Socket.IO, and Monaco, enabling conflict-free document editing across multiple clients.
• Designed a durable PostgreSQL state layer using an append-only update log and periodic snapshot compaction, supporting fast recovery without replaying complete document histories.
• Built the React/TypeScript client and NestJS backend, with Redis pub/sub coordinating updates across service instances.

• Implemented an Operational Transformation engine from first principles, supporting more than 30 concurrent users with synchronization latency below 150 milliseconds.
• Built a WebSocket synchronization layer with Redis pub/sub to distribute edits across Node.js server instances.
• Developed the React/Node.js/MongoDB application, which was used by 25 Harvard computer science students.

• Architected a Python/DSPy framework that injects, isolates, and repairs query- or answer-stage RAG failures without rerunning unaffected stages.
• Demonstrated across 300 HotpotQA examples that query repair improved exact match by 19.3 percentage points, while answer repair recovered only 2.2% of failures.
• Engineered interchangeable BM25/dense retrieval and OpenAI/Ollama backends with cost and latency telemetry, 36 deterministic tests, 92% targeted coverage, and automated CI.

January 2026 – May 2026

• Re-architected a synchronous generation pipeline as an asynchronous FastAPI worker system with task queues, parallelizing multi-stage inference and reducing per-video turnaround to under 90 seconds.
• Implemented a Redis-backed caching layer for reusable model responses, reducing API costs by 35% while supporting weekly use by more than 50 students.
• Owned backend architecture, caching strategy, and deployment, transforming a research prototype into a production service used by the club each week.

September 2025 – December 2025

• Built and deployed a real-time anomaly detection pipeline on AWS using Python, processing more than 10,000 security alerts daily with under three-second latency and less than $100 per month in cloud costs.
• Engineered Pandas/NumPy feature pipelines and trained scikit-learn models, achieving over 80% precision while reducing false positives by 60%.

May 2025 – August 2025

• Developed a backend reconciliation service using REST API integrations and SQL-based status checks, automating matching across more than 50,000 daily transactions and reducing reconciliation time by 35%.
• Implemented custom matching, validation, and exception-reporting logic to identify failed, delayed, and mismatched payments before settlement.
• Increased reconciliation accuracy to over 99%, reducing operational risk across high-volume payment settlement workflows.

June 2024 – August 2024 · Cape Town, South Africa

• Developed Python data pipelines to clean, standardize, and transform unstructured labor records into analysis-ready datasets.
• Produced reliable datasets that supported client-facing analysis, stakeholder presentations, and policy recommendations.

September 2023 – Present

• Build and maintain a full-stack learning platform supporting student onboarding, content delivery, and program operations using React, Node.js, TypeScript, GraphQL, and MongoDB.
• Translate staff and student workflows into product improvements while coordinating the academy’s day-to-day program operations.