AI engineering R&D studio · building in the open
Building the systems that make AI software more reliable.
Suiflex is an independent AI engineering research and development studio. We explore the engineering around AI agents—infrastructure, evaluation, testing, and developer tools—then turn useful ideas into open-source systems you can run, inspect, and improve.
Our process is simple:
research → experiment → build → release → learn
We focus on practical questions that recur when AI systems become part of real engineering workflows:
- How can AI agents work with more discipline, less waste, and clearer evidence?
- How can testing and verification remain rigorous as systems become more autonomous?
- Which focused tools make AI-assisted engineering simpler without hiding the system underneath?
- Where do fast, portable, user-controlled tools make a better alternative?
AI engineering discipline — a token-efficient quality layer for AI coding agents.
One Rust binary turns explicit engineering intent into checks, tasks, and evidence across coding agents—without adding another LLM call to the loop.
Developer tooling — a free, open-source desktop database client built as one native binary.
A cross-platform Rust and Slint application for PostgreSQL, MySQL, Redis, MongoDB, SQLite, Cassandra, SQL Server, and ClickHouse.
Agent-native testing — a self-hostable, MCP-native QA platform for deterministic and agent-driven testing.
Suitest combines black-box DOM testing, video and screenshot evidence, MCP workflows, and optional local or cloud models.
Not every repository starts as a product. Some are public experiments, infrastructure, or questions we are working through in the open:
- companion — a public supporting repository in the Suiflex ecosystem.
- websift — an early public experiment.
- homebrew-tap — Homebrew distribution infrastructure for Suiflex binaries.
- scoop-bucket — Windows distribution infrastructure for Suiflex binaries.
- Build executable research, not just research-shaped documents.
- Measure the system before making the claim.
- Keep users in control of their tools, models, and data.
- Release the useful parts in the open whenever practical.
You do not need permission to fix a small, verified problem. Larger changes start with a conversation so effort lands where it matters.
- Fix what you can reproduce. Show the failing case and keep pull requests focused.
- Propose before you build. For new behavior or architecture, open an issue first.
- Use it under real pressure. Run the tools in your own workflow and report what breaks.
- Badrus Shoolehk — co-founder · software architecture
- Mulham — co-founder · product & engineering
We are looking for founding partners who want durable, open developer infrastructure to exist. Support funds engineering time, release infrastructure, and the maintenance that keeps public tools useful—without buying control of the roadmap or a hidden private fork.
Discuss a founding partnership →
© 2026 Suiflex
