Hi, I’m Nithilan. I study Computer Engineering at the University of Wisconsin–Madison and build games, engines, and the occasional dev tool. I’m currently a gameplay engineering intern at Neureal, working on a Roblox project called SKAR. I write about game dev and AI on Medium and sculpt dark fantasy 3D models on Sketchfab.
I work where systems programming meets play. Most of what I make starts as a question about how something works underneath, a renderer, a narrative engine, a search algorithm, and turns into something you can actually pick up and run.1
My background is split between low-level C++ (engines, roguelikes, custom memory management) and shipping gameplay features on live platforms like Roblox. Lately I’ve been pulled toward on-device AI, getting real models to run fast on hardware that fits in your pocket.
“Build the thing that teaches you how the thing works.” – operating principle
Gameplay Engineering Intern – NeurealNow
Building gameplay systems for SKAR, a Roblox project, Lua systems work, in-game economy, and tooling for the design team.
Software Engineering Intern – Sentari AI
Worked across an AI product stack, model integration, latency-sensitive inference paths, and the glue that turns research into something shippable.
Software Systems Intern – Carrier
Software for connected systems, embedded-adjacent work, data pipelines, and internal tooling at industrial scale.
Software Engineering Intern – Neureal
Second tour, deeper ownership of gameplay features and live-game systems.
Software Engineering Intern – Neureal
First role in game development, learned the production pipeline and shipped features to a live audience.
A C++ roguelike built around two systems: procedural dungeon generation, and a DAG-based branching narrative engine where story beats unlock along the edges of a directed acyclic graph. The same node can be reached by different paths, but never twice, so the narrative stays coherent while the dungeon underneath it stays unpredictable.
A custom 2D engine written from scratch in C++ on top of SDL2. It owns the full loop, input, an ECS-style entity model, real-time collision, rendering, and a custom memory manager so allocations don’t fragment the heap mid-frame. The goal was to understand every layer between the keyboard and the pixel.
A local-first smart-glasses AI assistant: voice in, vision understanding, speech out, the entire round trip in under 800 ms on a MacBook Air, using only on-device models. No cloud, no API calls. Speech via distil-whisper, vision via Moondream 2, and Piper for TTS, with the latency budget split so nothing dominates the frame.
An in-game shop for the Roblox title Space Knights, with fuzzy search across hundreds of weapon items. Typos still find the right gun: a Levenshtein-distance ranking scores every candidate against what the player typed, so “exalibr” still surfaces Excalibur at the top of the list.
Notes on game development, systems, and on-device AI, published on Medium as @Sathariels.
The best way to reach me is email. I’m always happy to talk about engines, roguelikes, or getting models to run where they shouldn’t.