Kevin McLeod
Software Engineer @ Harvard
Hi! I'm currently a junior at Harvard pursuing a double major in Computer Science and Physics. I'm passionate about all different kinds of software and engineering, from AI to web development to robotics to game development! In my free time, I love dancing, playing instruments, running, playing frisbee, travelling, trying new restaurants and more!
Experience
AI Stealth Startup
- Co-founded an AI startup; conducted market research, product strategy, and user surveys of 100+ early users to align technical development with customer needs
- Led product design and full-stack development (React/Next.js/Tailwind, backend + AI models), delivering a modern web app adopted by early users
- Established agile processes for rapid iteration and oversaw product lifecycle from ideation through deployment, coordinating design, engineering, and user feedback
GP Enterprise Solutions
- Fixed critical platform bugs and optimized workflows, reducing user-reported errors by 40% and improving retention
- Restructured SQL queries and filter logic, cutting database response times by 20% and enabling faster access to high-volume financial data
- Shipped high-demand features (file downloads, form management) and collaborated across teams to refine APIs and frontend UX
Harvard AI Research Lab
- Developed a PyTorch pipeline combining Mask R-CNN and MiDaS to extract per-object 3D centroids from RGB data
- Designed energy-based compositional models to evaluate and optimize scene plausibility for embodied AI agents
- Benchmarked performance on novel indoor layouts and released open-source tooling to support reproducible research
Harvard Datamatch
- Redesigned the Datamatch UI in Next.js with a cross-functional team, improving clarity and UX for 22,000+ annual users
- Implemented new features for event promotion and user engagement while maintaining a consistent visual identity and brand tone
FIRST Robotics Team 1493
- Led electrical and programming subteams, building competition-ready robot control systems and reliable codebases
- Directed pit operations at high-stakes competitions, adaptively troubleshooting hardware/software issues under strict time constraints and pressure
- Placed 23rd at the 2022 FIRST Robotics World Championship (Houston, TX); organized STEM outreach programs and community initiatives
- Built an automated computer vision aiming system that aligned the shooter in real time, boosting scoring accuracy 90% and reducing the need for manual adjustments during matches
Projects
A few things I’ve shipped recently.
Personal Project
Created a Python/Pandas + Google APIs platform to compare 4,000+ colleges across hundreds of criteria. Enabled real-time filtering, multi-dimensional search, and automated updates through integrated APIs.
Harvard Negotiation Club
Launched the Harvard Undergraduate Negotiation Club’s first website with React/Next.js. Delivered a mobile-friendly platform with event announcements, member bios, and recruitment resources.
Academic research in computer vision and robotics
Researched modular 3D scene understanding using energy-based models for embodied agents. Built PyTorch pipelines combining Mask R-CNN and MiDaS to extract object-level 3D centroids and optimize scene plausibility.
Campus matchmaking & event platform
Redesigned Datamatch’s website in Next.js with a cross-functional team, improving clarity and UX for 22,000+ annual users. Enhanced accessibility and performance while adding event features to boost engagement.
Personal Group Project
Directed a 10-person team to design and build a 3D RPG game with Unity. Developed modular combat, dialogue, cutscene, and animation systems, ensuring smooth gameplay and cross-functional collaboration.
Personal Project/Hackathon
Developed a Unity game blending Space Invaders, Breakout, and Flappy Bird mechanics. Players shoot meteors to stay afloat via recoil, balancing survival and offense in a retro 2D-in-3D style.
OS Class
Built a minimal OS in C++ with a custom heap allocator, virtual memory, and syscall interface. Added thread-safe I/O and a preemptive scheduler to simulate multitasking and safe concurrent file access.
Competition Entry
Led a 3-person team to build an agent for Correlation One’s Tower Defense competition. Ranked 75th of 66,265 players and achieved a 60% win rate across 7,425 matches with adaptive offensive/defensive strategies.
Tech for Social Good Club Project
Built a React/Next.js + Supabase web app with fuzzy search, real-time updates, and adaptive UI. Designed responsive Tailwind interfaces to support mobile and low-latency use for biodiversity researchers.
Personal Project
Recreated the classic Chrome Dino game in Python with Pygame. Implemented obstacle spawning, collision detection, and increasing difficulty for an endless runner experience.