Applied AI Engineer
Summary
Build and ship production trading systems using AI-first development tools like Claude Code and Cursor, focusing on rapid prototyping and autonomous AI agents.
The challenge
We're not hiring software engineers who sometimes use AI. We're hiring builders who don't write code without it.We need you to ship applications, platforms, and systems using Claude Code, Cursor, and whatever comes next as your primary development environment. The underlying AI knowledge isn't the point. Building is.
Why this matters
AI-first development isn't a productivity hack. It's a different way of building software. You design systems by describing them. You refactor by explaining what's wrong. You debug by showing the AI the error and the context. You ship in days what used to take weeks.
We've proved this works. Whether it's structured spec-to-implementation workflows or conversational vibe coding with frontier models, we ship production systems using AI as the primary development interface. Now we're scaling it. We need engineers who've already made this shift.
Why Deriv
Our mission is Trading for Anyone, Anywhere, Anytime. Millions of traders, around the clock, across regulatory regimes. We need to ship faster than traditional development allows.
We're shipping, not experimenting. We build products for over 3 million traders, built with AI and live in production. Amy handles all customer enquiries: ~80% resolved autonomously, the rest are resolved through agent-to-human collaboration on Slack. We're ideating and building entirely new trading experiences with AI.
We're building autonomous systems to run entire business functions. HR, security, marketing, compliance, product and finance. We're creating systems that operate these functions with human collaboration. This is the direction we're heading.
We share what we learn. Deriv documents what we're shipping, what breaks, and what we figured out the hard way.
What you'll do
- Ideate and build with AI. Use AI to explore problem spaces, generate product ideas, and define what to build. Go from vague problem to working prototype without waiting for detailed specs
- Code with AI as your co-developer. Use Claude Code, Cursor, or similar tools as your main interface for building. Generate, refactor, test, and document code through AI collaboration. Evolve your workflows continuously
- Review and improve, not just generate. Use AI to review AI-generated code, catching issues at scale. Build guardrails and feedback loops that surface problems automatically
- Ship end-to-end. Own the full arc—idea, architecture, implementation, deployment, iteration. Ship features that go to production. Not prototypes. Not demos
- Explore autonomous AI systems. Help build and deploy AI agents that run continuously, maintain context, take real actions, and work proactively. This is experimental territory—you'll help define how it works
Who you are
- You already code this way. 2-8 years of software engineering experience, and you've made the shift to AI-first development. You reach for Claude Code or Cursor before you reach for the keyboard. This isn't aspirational for you; it's how you work.
- You ship production software. Not POCs or side projects that never launched. Real systems that handle real users, break in unexpected ways, and taught you things the AI didn't know.
- You move fast and fix things. Speed matters. Perfect is the enemy of shipped. You prototype, test with real users, and iterate based on feedback. When something breaks, you fix it. You don't file a ticket.
- You're a builder, not a specialist. Full-stack comfort. Frontend, backend, APIs, databases, deployment. You do what the project needs. You're not precious about your domain.
- Bonus: You've experimented with autonomous agents. Claude Agent SDK, OpenClaw, OpenAI Codex agents, or built your own always-on AI systems
Tech stack
This role is tool-agnostic. What matters is output, not allegiance to specific models or IDEs.
Current tools we use:
- AI Development: Claude Code, Cursor, Claude (Opus, Sonnet), GPT-5.2, Grok 4.1, Qwen3, GLM-4.7, Kimi K2.5
- Autonomous Agents: Custom agent frameworks, experimenting with tools like OpenClaw
- Languages: TypeScript, Python, and whatever the project needs
- Infrastructure: AWS, GCP, PostgreSQL, Redis, Docker
The honest reality
This role demands speed. You'll be expected to ship in days what traditional teams ship in sprints. You'll prototype features before requirements are fully locked. You'll deploy to production more often than feels comfortable.
AI doesn't eliminate complexity—it accelerates it. You won't debug line-by-line. You'll use AI to find issues in AI-generated code, build feedback loops that catch mistakes automatically, and create self-learning systems that improve over time. The tools keep changing. The systems you build will need to learn and adapt alongside them.
But you'll ship more than you've ever shipped. You'll build systems that matter, not backlogs that grow. And you'll work with people who've stopped debating whether AI changes development and started proving it.
If you want well-defined sprints and stable toolchains, this isn't it. If you want to build fast and ship constantly, let's talk.
What’s good to include in your CV
- Evidence of AI-native development: GitHub repos, deployed projects, or examples of what you've built with Claude Code/Cursor
- Brief note: What's the most complex thing you've shipped using AI-assisted development? What broke, and what did you learn?
- Bonus: Any experiments with autonomous agents (Claude Agent SDK, Moltbot, custom setups, etc.). We'd love to see what you've tried