The role
We're looking for an AI Product Engineer who efficiently ships outcomes. You'll receive a user problem, a rough capability, or a direction like "expose this through MCP", not necessarily accompanied with a full interface design. Your job is to chart a path to the end-state, sequence the unknowns so you're never blocked on a single big bet, and put working functionality in the users' hands at every stage. We innovate and move fast. The focus is on thin vertical slices that compound to produce impressive features.
Concretely, that means extending Duet's agent harness: integrating new data sources, building MCP access to Duet functionality, and authoring curated prompts and skills. The AI Product Engineer needs to bring their expertise to these processes to implement robust and cutting-edge solutions with incremental wins along the way.
How we expect you to work
Start from outcomes, not design requirements. We are obsessed with delivering value to our customers. You translate user needs into solutions with a staged delivery plan that brings value to the user in each sprint.
Be curious and engaged. We are a small company that values the voice of every team member. We expect every engineer to become familiar with the world of intellectual property and to bring your ideas and talents to the table so that we can create the best possible solutions for our customers.
Collaborate, challenge, and support your team members. We live the spirit of agile - we are radically transparent, we take introspection seriously, and we constantly adapt and improve. You will be joining a highly collaborative team that thrives on exchanging ideas, challenging each other, and learning from each other.
What you'll work on
MCP server access
Design and implement our approach to expose Duet's capabilities via MCP.
Rapid data integration
Add new data sources by writing thin TypeScript connectors and wiring them into the agent as composable tools. You'll often get "we need source X in the agent" before every field and failure mode is defined. Take the lead to understand the data source and find the most efficient way to incorporate it into Duet.
Curated prompts & skills
Author and iterate on system prompts, tool designs, skill definitions, and the architecture that links these resources together.
Agent platform work
Extend and improve our production agent loop. You'll work on context management, streaming, Redis locks, billing, presentation to the user, etc.
What we're looking for
- 3+ years professional software engineering experience, with strong TypeScript/Node backend skills
- 12+ months building LLM agents, with an emphasis on integrating orchestrating complex tools and data sources
- A track record of delivering outcomes efficiently
- Incremental delivery discipline - you decompose work so that each PR adds user value or represents a deliberate milestone that leads to the next piece of functionality
- Demonstrated prompt engineering as product work - structured prompts, iteration from failure modes, version control
- Experience integrating third-party APIs into an agent framework
- Excellent written and verbal communication with a focus on teamwork and collaboration
- mphasis on integrating orchestrating complex tools and data sources
- A track record of delivering outcomes efficiently
- Incremental delivery discipline - you decompose work so that each PR adds user value or represents a deliberate milestone that leads to the next piece of functionality
- Demonstrated prompt engineering as product work - structured prompts, iteration from failure modes, version control
- Experience integrating third-party APIs into an agent framework
- Excellent written and verbal communication with a focus on teamwork and collaboration
