Computer Engineering student and Interviewlary co-founder building applied-AI and full-stack systems across LLM infrastructure, voice AI, agent workflows, and production web products. Strongest in product definition, integration, debugging, testing, and evidence-based release QA.
Co-founded and helped ship a live AI interview-training SaaS with role-specific question generation, bilingual voice/text mock interviews, transcription, scoring, feedback, subscriptions, and performance history.
Lead product requirements, provider evaluation, technical QA, and release review across LLM generation, STT/TTS, authentication, billing, scoring, feedback, latency, and failure-path behavior.
Contribute to implementation and debugging through repository inspection, logs and Sentry, integration work, journey testing, and edge-case verification; translate product failures into scoped engineering fixes.
Technical projects & systems
TokensCache
Shipped · Public
TypeScript and Node multi-provider LLM optimization layer with exact and semantic caching, prompt optimization, provider adapters, real-dollar budget controls, and MCP tooling. 92/92 tests passing; 35.5% fewer billed tokens in repository mock A/B scenarios.
RoutineAI
Live · Active build
Next.js and Supabase student-day OS with conversational setup, deterministic scheduling, reminders, focus tools, ICS export, web push, and task continuity. Passed 106/106 tests plus production build, lint, typecheck, and secret scan.
TruthNet
Public · AI hackathon
FastAPI and React real-time fact-checking system using a four-agent adversarial pipeline for evidence gathering, prosecution, defense, and final judgment. Built and presented as a working product in a seven-hour hackathon sprint.
Lary OS
Active build
Interviewlary founder command center with AI briefings, Railway, Supabase and Paddle monitoring, provider cost tracking, scheduled email digests, signed reporting, and audited actions; releases 0–4 implemented in code.