I’m a software engineer and MSCS student at Northeastern University in Silicon Valley. I build full-stack applications and AI systems, connecting backend infrastructure with interactive experiences.
My interests include human–AI interaction, multimodal systems, and making AI behavior more reliable and measurable.
LinkedIn · VOICE · CueFrame demo
| Started | Project | What it does · Explore |
|---|---|---|
| May 2025 | VOICE Applied AI · Research project |
Multi-agent virtual-patient platform deployed in a graduate speech-language pathology (SLP) course in Summer 2026. I lead backend engineering and LLM evaluation, connecting interactive patient dialogue with automated assessment. Recognized with Northeastern University’s 2025–2026 Bridge Builder Award for this work. Platform · Code |
| Sep 2025 | Momentum Full-stack mentorship · Next.js / PostgreSQL |
Full-stack mentorship platform for goal setting, progress tracking, and mentor feedback. This project was used in Northeastern University’s OGS Global Student Mentor Program, supporting 40+ students. Live demo · Code · Architecture |
| Mar 2026 | CreatorPal Audience research · Google ADK |
Audience research agent built with Google ADK, combining hybrid retrieval, on-demand skills, and executable Python analysis. Checks numerical claims against computed results before committing recommendations with supporting evidence. Architecture · Evaluation |
| Jul 2026 | Agent Reliability Harness Agent evaluation · Python |
Fault-injection and code-validation framework for agent systems. Independently verifies database and file state, tests crash recovery and idempotency, and gates code patches through five isolated Docker checks. Recovery example · Report |
| Aug 2026 | CueFrame Visual interaction · C++ |
C++ visual-interaction runtime with local object detection, temporal evidence tracking, and rejection of outdated responses. Latest-frame scheduling reduced p95 observation latency by 44% versus FIFO in a controlled CPU benchmark. Interactive demo · Architecture · Benchmark |
| Sep 2026 | TurnServe Inference systems · C++ |
C++20 inference runtime for concurrent conversations sharing a llama.cpp model. Three scheduling policies coordinate prefill and decoding; cancellation-safe turn replacement prevents further output and history commits after cancellation is acknowledged. Design · Recorded run |
I enjoy traveling, working out, and swimming. I’m also a skyscraper enthusiast with an interest in architecture.



