Zachary Martin

Graphics concentration

Selected courseworkObject-Oriented Systems Design · Algorithms · Probability for Computer Science

• Developed the customer support dashboard with AI ticket ingestion from Gmail. The workflow uses customer/billing data, known issues, GitHub MCP, and Axiom logs to diagnose reports, draft replies for review, and group related tickets in an engineering queue.
• Delivered the Agora live coaching dashboard with booking-linked access, pre-join device setup, screen sharing, chat, and session tracking.
• Implemented Zacal for coach availability, paid bookings, cancellations, and rescheduling across calendars and time zones, with automated booking emails and atomic reservations to prevent double bookings.
• Created Dashboard Mod to manage dashboard roles and feature permissions, with bulk grants/revocations, server-verified authorization, and permission-change audit history.
• Developed Med+ onboarding flows for eligibility, intake, consent, and checkout, integrating Tellescope and OpenLoop for clinical-system handoff, with saved progress and recovery for interrupted sessions.

• Led design and development of a skincare website for product recommendations, client plans, and routine tracking; also worked on an iOS app.
• Built a RAG chatbot pipeline, AI-assisted routine tracking, LLM image analysis, and YOLOv8 acne detection.

• Taught Java, Python, and problem solving to beginner and intermediate students preparing for AP Computer Science and USACO.

• Developed a React/TypeScript design editor for reusable components, auto layout, and interactive prototypes, with Liveblocks collaboration and Supabase persistence.
• Added connected-board prototypes and flow validation to identify missing screen states, unreachable frames, and broken destinations.
• Enabled teams to review and merge design branches, publish shared libraries, resolve conflicts, and recover offline edits.
• Integrated an LLM design agent that creates and edits canvas objects through 180 typed actions, with scoped changes, undo that preserves human edits, and spending limits.

• Implemented a multithreaded C++20 path tracer with BVH acceleration, imported meshes, reflective and refractive materials, caustic photon mapping, and a progressive viewer.
• Trained a PyTorch U-Net and integrated ONNX inference into C++; averaged 35.96 dB PSNR versus 25.84 dB for raw rendering across 360 test images from five held-out layouts.
• Created reproducible training datasets with independent reference renders and scene-level splits, then extended the pipeline to joint denoising and 2× reconstruction.

• Three NCAA team championships (2021-2023); three NCAA All-American honors (2020-2022); MPSF vault champion (2023); two-time CGA All-America Scholar-Athlete.