Junior Developer - AI-Augmented Product & Tooling Team

USD Pay · Long-Term · Philippines

What We're Building

We build AI-powered consumer products, custom developer tooling, and client solutions across multiple verticals:
AI image and video generation — a mobile and web app that transforms photos into stylized artwork using generative AI. Single Flutter codebase, Firebase backend, multiple AI model integrations.

Game development — interactive experiences with structured world-building and AI-assisted content generation.

Prompt research and skills tooling — we built a package manager for AI agent prompts and skills that detects your repo's tech stack and ranks results accordingly. We research how prompts perform across different models, build reusable skill packages that any AI coding agent can load, and maintain a catalog of 30+ skills covering everything from security review to prompt quality self-assessment. This isn't a side project — it's a core product.

MCP and agent workflow research — we actively research and build integrations using the Model Context Protocol (MCP) and other emerging standards that reduce friction between developers and AI agents. We've built coordination servers that let multiple agents reserve files, message each other, and avoid conflicts. We've built tmux-based orchestration so developers can run 3-5 AI agents in parallel. When a new protocol, tool, or workflow pattern emerges, we evaluate it against real work and adopt or build what's missing.

Code quality tooling for agentic development — AI agents write a lot of code fast. The bottleneck isn't speed, it's confidence. We built compiler-integrated mutation testing and code quality metrics in Rust and TypeScript — tools that use each language's native AST parser so agents and developers can trust that generated code actually works. These aren't standalone research projects; they're the quality backbone that makes high-velocity agentic development safe.
Infrastructure and DevOps — Kubernetes clusters, CI/CD pipelines, private networking, infrastructure-as-code across all projects.

How We Work
We follow a pattern that defines the team culture: see a gap ? research what's available ? build it or adopt the best option ? ship it ? share it with the team.
We needed mutation testing tuned to our languages — generic linters couldn't handle Rust's error propagation or TypeScript's union types correctly. So we built compiler-integrated tools that use each language's native AST parser. We needed AI coding agents to coordinate without stepping on each other's files. So we built an MCP coordination server and a tmux orchestrator. We needed an issue tracker that works inside AI agent sessions. So we ported an open-source one to Rust and integrated it into the workflow. We needed to understand why some prompts waste tokens and others don't. So we built a prompt quality analyzer that extracts your messages from session logs, rates them, and shows you exactly where you're losing efficiency.

A new AI model drops on OpenRouter — by the end of the week, someone on the team has tested it against our workflow and reported back. A new MCP server appears on GitHub — someone's already evaluated whether it solves a real problem for us. A new ---------- tool shows up on Twitter — someone's tried it in a real session. This isn't a scheduled R&D activity. It's how the team operates.

The internal tools are mostly Rust and Go. The products are Flutter and TypeScript. Client work varies — we recently picked up a Drupal project, and the expectation was that the team could ramp on an unfamiliar framework in a day using AI-assisted exploration. That's the skill we hire for.

What Makes This Team Different
Every developer works alongside AI coding agents as core workflow — not as an experiment. You'll have:
5 AI coding runtimes at your workstation (Claude Code, Codex CLI, Gemini CLI, OpenCode, Cursor)
30+ custom agent skills covering security, testing, architecture, code review, and prompt quality analysis
Multi-agent coordination via MCP so you can run several AI assistants on different tasks simultaneously without file conflicts
Automated quality gates (mutation testing, code metrics, security scanning, bug scanning) that give you confidence in AI-generated code before it hits review

Prompt research tools that help you measure and improve how effectively you communicate with AI agents
We don't just use these tools. We build them, research the protocols behind them, improve them, and expect every developer to have opinions about what works and what doesn't.
What We Expect From You

You experiment before you're told to. When a new AI model, CLI tool, MCP server, or workflow appears — on Twitter, Hacker News, a ---------- , wherever — your instinct is to install it and try it. Not bookmark it. Not wait for a tea ---------- eting. Try it, form a view, share what you found.

You assess your own work. We have tools that analyze prompt quality, code review effectiveness, and development patterns. You're expected to run them on yourself, identify your own anti-patterns, and fix them. Nobody will chase you to do this.
You learn by doing, not by reading. Our onboarding gives you access to the full environment on day one. We expect you to explore it. Open things, break things, ask questions when you're stuck — but try first. If we hand you a framework you've never seen before, the expectation is that you and your AI agents figure it out together — fast.

You communicate early, not late. Blocked for 30 minutes? Say something. Found a problem nobody mentioned? Flag it. A process doesn't make sense? Push back. Silence is the only wrong answer.

You follow through. When you pick up a task, you own it until it's done — including tests, docs, and letting the team know it's complete.

What the Work Looks Like
Build and ship features in Flutter (Dart) and Firebase Functions (TypeScript)
Contribute to internal Rust and Go tooling as you grow into the stack
Research and evaluate new AI models, MCP integrations, and developer tools
Integrate AI models for image generation, video generation, and content moderation
Pick up unfamiliar frameworks for client projects using AI-assisted exploration
Conduct peer reviews that actually catch problems — not rubber stamps
Evaluate new tools, models, and workflows as they emerge and share findings with the team
Run self-assessments on your development practices and discuss results at standups
Improve the team's engineering standards — our docs are living documents, not commandments

Tech Stack
Products:
Flutter (Dart) — iOS, Android, web from one codebase
Firebase Functions Gen 2, Node.js 25, TypeScript
Firestore, Cloud Storage, Stripe
OpenAI API, Google Gemini, Vision API

Internal Tooling:
Rust — mutation testing, code quality metrics, issue tracking, bug scanning
Go — AI skill package management, multi-agent orchestration, MCP coordination servers
Infrastructure:
Kubernetes, Forgejo (self-hosted git), Jenkins CI/CD
Ansible, OpenTofu, Semaphore, SAST/DAST and other security tooling
Lefthook, oxlint, Prettier, Spectral, Checkov, gitleaks, semgrep

You don't need to know all of this on day one. You need to be the kind of person who starts exploring it on day one.
Schedule

The team currently works 4pm–12am PHT (Manila time), but other schedules overlapping with US Eastern business hours can also be accommodated. We work this window as a team. As we grow, we plan to introduce flexible scheduling as additional teams are creating with other schedules.

Compensation
We're a bootstrapped startup pre-launch, looking for junior developers who are hungry to build real skills on real products. Starting compensation is junior-rate USD, paid via OnlineJobs.ph / Payoneer. Compensation scales once we launch and have revenue — early tea ---------- mbers who prove themselves will be first in line for increases.

What you get:
Production experience shipping AI consumer products with professional engineering practices
Daily hands-on experience with AI coding tools and internal Rust/Go tooling most developers never touch
Exposure to prompt engineering research, MCP integrations, and agent workflow design
Direct mentorship on a small team — no layers between you and decisions
USD income with growth tied directly to company growth
A real voice in how the team works and what tools we build next

How to Apply
Step 1: Pick any AI coding tool you haven't used before — Claude Code, Codex CLI, Gemini CLI, RooCode, Kiro, GitHub Copilot, anything. Spend 30 minutes using it on a small task (a personal project, a tutorial, a three.js game, a port of a popular utility into Rust, a throwaway script — doesn't matter).
Step 2: Send your application with:
What you built with the AI tool — what you tried, what surprised you, whether you'd use it again. A few sentences is fine. We don't care which tool or whether you liked it. We care that you tried it and formed a view.
A project you've built that you're proud of (link or description)
Your experience with Flutter, TypeScript, Rust or similar
Your availability (full-time or part-time)
Your expected monthly rate (USD)