MS Computer Science @ Arizona State University (May 2027) · Software engineer working across backend systems and applied AI.
I like building things end to end: typed backend services with real test suites, LLM agents that have to be right rather than just fluent, and the evaluation work that tells you which one you've actually got.
🎯 Open to new-grad 2027 SWE / AI-ML roles
Software Engineering Intern · Optiwise (AgriChoice) · Jun – Aug 2026 Farm-equipment rental marketplace. Shipped a rental-extension billing feature end to end across four TypeScript/Fastify microservices and a Kotlin/Jetpack Compose Android client, with GST-compliant documents in four languages. The backend runs ~2,400 automated tests and uses event-driven messaging (pg-boss) on PostgreSQL/PostGIS.
Graduate Service Assistant · ASU School of ECEE · May 2026 – Present ML Research Intern · Interlace India · Aug – Sep 2023: real-time email-classification pipeline in Python
| Project | What it is | Stack |
|---|---|---|
| DispatchIQ | AI fleet command center (GlobeHack S1, Trucker Path track). LLM agents score and recommend drivers from live GPS, HOS and cost data. A Groq Vision OCR pipeline cut invoice processing from 45 min → 90 s. | Next.js · TypeScript · Groq · Twilio |
| Roomies | Full-stack apartment & roommate manager covering expense splitting, rent, chores, maintenance, presence and group chat | JavaScript · Node.js |
| COMPAS Fairness Audit | Audits a criminal-risk model for bias across race and gender, checks calibration, and routes sensitive cases to human review | Next.js · FastAPI · Python |
| LinkedIn Post Generator | Few-shot LLM app with tone, length, language and audience controls | Python · LangChain · Groq · Streamlit |
| Where Did the Forest Go? | Maps Amazon deforestation from free Sentinel-2 satellite images with a U-Net. Hand-checked on 600 blind, stratified points; switching the training labels to MapBiomas raised precision for cleared forest from 59% to 88% (92% overall accuracy), estimating about 6,800 ha cleared from 2021 to 2026 | Python · PyTorch · GeoPandas · Folium |
| MeloLearn | Music-learning platform with audio editing tools and a content-ranking leaderboard | JavaScript · Express.js |
Research & coursework
- Controlled Persistent Memory for LLM Assistants: co-authored a paper on a write-routed, versioned memory architecture. Evaluated on 588 PersonaMem episodes, it reached 66% accuracy while storing ~60% fewer memories than vanilla RAG.
- Clickless: a three-agent Planner → Retriever/Ranker → Explainer shopping assistant over 25–30 GB of Amazon product data, reaching 92% precision@50.
- 📄 OTP Verification & Zero Trust Security for RBAC in Medical Records, ICSTSN 2024
Languages: Python · TypeScript · JavaScript · SQL · Kotlin Backend: Node.js · Fastify · Express · Prisma · PostgreSQL/PostGIS · Docker · REST · event-driven systems AI/ML: LLM agents · RAG · LangChain · evaluation harnesses · PyTorch · scikit-learn · OpenCV Frontend: React / Next.js · Jetpack Compose · Streamlit
