experience

view resume ↗

Data Analytics Fellow (Analyst, IT) — Duke University

Building production data workflows, dimensional warehouse models, and dashboards for Duke and Duke Health research.

SQLDimensional modelingETLDashboards

Founding Software Engineer, Backend — Aquara.ai

Building the backend for an AWS-to-Azure database migration, and the authentication flows on top of it.

AWSAzureAuthenticationBackend

Software Engineer Intern — BTG Pactual Bank

Built and shipped “Log in with BTG ID”, a production distributed SSO service giving 6,000+ employees and 8,000+ small-business clients one authentication path across BTG Pay products. Cut authentication latency 47% toward a 1-second target by caching session state in Redis, and personalized content on EXAME, BTG's news platform, by connecting GraphQL banking-profile data to the OAuth 2.0 flow. Owned rollout and reliability — authored the RFC, shipped behind feature flags, and restored service within SLA leading incident triage with live telemetry and a fallback path.

OAuth 2.0OIDCPKCETypeScriptNode.jsGraphQLRedis

Technical Assistant, Prototyping & Fabrication — Duke OIT Innovation Co-Lab

Cut 3D-printing turnaround from 3+ to 1–2 business days by managing end-to-end delivery for 24 stakeholders. Held client satisfaction at 95% while producing 47 custom models for Duke Health research and standardizing production specifications.

PrototypingFabricationProcess design

Software Engineering and AI Mentee — Microsoft and Persistent Systems

Cut healthy-food search time 85% — from 5 minutes to 45 seconds, at 80% task success — by building MealPilot with Microsoft Copilot Studio and the Azure Maps API. Selected as 1 of 18 students nationwide for the Microsoft NYC Summer Mentorship Program and AI Hackathon.

Copilot StudioAzure Maps APIPythonAI agents

Software Engineer Intern (Code+ Program) — Duke University

Saved Duke Facilities 2 hours a day on data requests by deploying a dashboard backed by a REST API, Docker, Linux and Postgres. Reached 95% precision on three-class occupancy classification and a cross-validated R² of 0.72 on regression, training scikit-learn models on 2M+ anonymized Wi-Fi and CO₂ records, then built WCAG-aligned visualizations with Facilities across 4 Agile sprints to support HVAC planning.

DockerLinuxPostgreSQLRESTPythonScikit-Learn