AI Support Engineer at Dust building AI agents and automation workflows to reduce manual support load. Role splits between running support for complex issues and engineering AI systems, agents, and tooling to eliminate categories of tickets. Requires fluency with Claude Code, Cursor, and agentic platforms.
Work is being rewritten, and the people holding the pen are the ones who actually run it.
We call them AI operators: the employees inside companies who build, deploy, and run AI agents for their teams, without waiting for someone to hand them a tool. Dust is the platform they choose to rewire how their company works.
With 70%+ weekly active users, people stick with Dust as much as they do with Slack and Notion. We don't get piloted and shelved. We land once, and spread. We're at an exciting stage of our journey, and growing fast.
We're serving great customers like Datadog, 1Password, Cursor, Clay, and Persona, and aim to x5 our growth by the end of 2026.
Dust is backed by Sequoia with a determined team of optimists (coming from Stripe, OpenAI, and Stanford) who like to focus on users, ship fast, and don't take themselves too seriously while doing so. The Generalist named us among the Future 50.
At Dust, we're coining the term AI Operator: someone who rethinks and rebuilds company processes around AI. Not "how can AI help us do this faster?" but "if AI existed from day one, would we even do this the same way?"
The AI Support Engineer applies this mindset to Support. You’ll split your time between running support (handling complex issues when agents fall short) and building the AI systems that reduce that workload over time. Support at Dust is not a cost center, it’s a product. Success is measured by eliminating categories of tickets, not just resolving them.
You will define, ship, and continuously iterate on the infrastructure that lets Dust deliver a world-class support experience at scale: AI agents, automation workflows, classification systems, knowledge systems, and tooling built on top of Dust itself. You will dogfood the product harder than almost anyone at the company.
Build (50%)
Design, ship, and maintain AI agents and automation workflows that reduce manual support load: think ticket classification, acknowledgment automation, response drafting, incident detection, and proactive user outreach.
Identify entire categories of issues and engineer them out of existence: better documentation, tighter agent prompts, or upstream product feedback.
Build and maintain tooling (MCP integrations, Dust agents, internal scripts) that increase the team's capacity without increasing headcount.
Contribute to the evolving "Support as a Product" backlog. You will own tasks with shipped/doing/todo statuses like a product engineer, not a queue operator.
Run (50%)
Investigate issues across logs, code, and internal tooling to identify root causes and provide clear answers.
Handle complex, escalated cases with precise and accessible communication
Systematically analyze agent-generated responses for inconsistencies and ambiguities. Iterate on prompts, documentation, and tooling until human intervention is minimal.
When the agent fails due to missing or incorrect information, close the gap — own the follow-up across engineering, product, and documentation.
AI-native: You're a power user of AI coding tools like Claude Code, Cursor, and agent platforms like Dust — this is how you work, not a supplement to how you work. You've built workflows, automations, or tooling with AI, and you can speak concretely to the tradeoffs of agentic systems (speed vs. accuracy, human-in vs. human-out-of-the-loop, deflection vs. resolution).
Technical debugging capability: Comfortable reading code, analyzing logs, and navigating a codebase to form a hypothesis. Full engineering expertise is not required, but you can't be stopped by a stack trace.
Builder mindset: Track record of reducing repetitive work through systematic improvements → you get bored doing the same thing twice and build a system instead.
Excellent written communication: Clear, concise communication for both technical and non-technical audiences. (English required for US based roles; French and English fluency required for Paris based roles).
Analytical precision: Able to identify patterns, edge cases, and inconsistencies others miss.
Collaborative persistence: Drives improvements across teams while maintaining strong working relationships.
High ownership: "Doer attitude" → you operate with autonomy, surface problems proactively, and drive outcomes without needing to be managed.
Curiosity and adaptability: Comfort with ambiguity and first-of-a-kind problems. This role does not yet have a defined playbook → you will write it.
This is a builder-first support role. You’re not just resolving issues, you’re designing the systems that prevent them.
You:
Default to building solutions instead of escalating
Use AI tools as your core work environment
Treat Support as a product with a backlog, not a queue
Focus on eliminating categories of work, not just completing them
Work fluently with concepts like deflection, agent loops, prompt design, and automation systems
If you've come from a company that treats support as strategic and you've spent time building support tooling rather than just using it, this role was written with you in mind.
Competitive compensation: €50K - €80K base salary - We can go higher for outstanding profiles.
We offer substantial support for relocation, including a stipend up to €10k, finding an apartment in Paris, and supporting you with all residence and work permit-related procedures.
Significant equity package in a Sequoia-backed startup
Health insurance for you and your dependents
New MacBook Pro or Linux machine, monitor, keyboard, etc.
Beautiful office in the heart of Paris
Opportunity to travel to the US multiple times a year
Regular team events and offsites
We're prioritizing building our team with an in-person culture at our offices in Paris and San Francisco, because we value the magic that happens when talented people work closely together.
The models are powerful enough. What's missing is the product layer where AI meets how companies actually work. That's what we're building: the infrastructure that lets any team turn scattered knowledge and tools into coordinated execution with agents they build, own, and run themselves.
We use Dust ourselves every day. We get to shape how humans and agents collaborate while solving our own problems with the product we ship. That loop is rare, and it's why we move fast.
If you're excited about defining a new category and want to join a determined team of optimists who focus on users, ship fast, and don't take themselves too seriously, we'd love to talk.
Even if you don't check every box in our requirements, we encourage you to apply. We value diverse perspectives and backgrounds, and we're more interested in your potential and passion than a perfect match to our checklist.
Learn how we think and work.
Our product constitution, a story about our mission
Agents at work - Latent Space, podcast with our cofounder, Stanislas Polu, 2024
LLMs reasoning and agentic capabilities over time - dotAI, podcast with our cofounder, Stanislas Polu, 2024