Flush Laboratory
  • Home
  • About
  • R&D Logs

On this page

  • Shipped Projects
    • GameJolt API Plugin — Unreal Engine 5
    • FlushTools — Single-Header C Utility Library
  • Active R&D
  • Technical Notebooks

FlushLabs

Systems engineering, engine tooling, and lightweight AI research.

I’m Flushwhy — a systems engineer with an HPC background, currently building engine tooling, networking utilities, and lightweight AI research for games.

This lab is where I ship open-source infrastructure, run benchmarks, and document the work publicly.

Field Technologies
Languages C/C++, Go, Odin, Python
Engines Unreal Engine, Godot
Domains HPC, Networking, Applied AI

Shipped Projects

GameJolt API Plugin — Unreal Engine 5

Native C++ · Subsystem Architecture · Blueprint Integration

A production-ready integration of the GameJolt API built natively for UE5. Bypasses bloated wrappers and plugs directly into Unreal’s subsystem architecture — exposing clean async pipelines for leaderboards, session management, cloud storage, and authentication with zero frame-budget impact.

View on GitHub


FlushTools — Single-Header C Utility Library

Bit Packing · Serialization · Quantization · Coordinate Compression · Networking

A lightweight, dependency-free C utility library built for engine-side use. The most useful parts are the networking and compression utilities — bit packing, quantization, and coordinate compression for multiplayer data optimization. Also includes a custom PRNG validated against 100M samples for uniform distribution and serial independence.

View on GitHub


Active R&D

Embedded Low-Power TTS

Deep Learning + Signal Processing

Ultra-efficient Text-to-Speech optimized for local inference on Raspberry Pi and mobile — no cloud round-trip. Replacing heavy APIs with native C/C++ inference to make runtime dialogue systems viable on constrained hardware.

→ R&D Logs

ML-Enhanced 2D Physics

Rigid Body Dynamics + Neural Approximation

A deterministic 2D physics engine combining classical constraint solvers with neural approximation for expensive dynamics — targeting sub-16.6ms execution to fit inside a real game loop.


Technical Notebooks

Benchmarks, statistical analysis, and implementation notes live in the R&D Logs. Everything is executable and reproducible.


NoteOpen to Collaboration

Looking to connect with systems engineers, game developers, and AI/ML researchers interested in real-time software. Commercial R&D, benchmarking, or early-stage venture building — reach out.

  • Discord: _flush
  • Signal: flush.01
GitHub Discord (_flush)