SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: CAD 250,000 - 280,000 / annual
Honeycomb is building Canvas, an agentic workspace where engineers investigate and understand their systems using AI agents that reason over high-cardinality observability data. The Agentic Intelligence team has shipped Canvas, the Honeycomb MCP server, Canvas Agent, and Canvas Skills surfaces. You'll expand what the team can build: new agents, new surface area in Canvas, memory, spatial awareness, improved performance on the Bedrock loop, and more.
Honeycomb's data store is fast and accepts high-cardinality data, enabling agents to do things no other observability product can—correlating across services, drilling into single traces, comparing before and after a deploy. This role leverages that unique capability to build production-grade agents that investigate, reason, and act on live observability data inside Canvas, trustworthy enough for engineers in high-pressure, mid-incident situations.
You will design and deliver production-grade agents, taking them from rough prototype through to something that holds up under production traffic. You'll own the agent work end-to-end—scoping, building, shipping, and maintaining agents including the evals that measure whether they improved. The role is agent-focused with some fullstack development. You'll extend Canvas, the MCP server, and Canvas Skills with new capability, make the case for what comes next with working code, and define what "good" means for agents at Honeycomb: measurable against real evals, maintainable, and honest about limits.
Example projects include multiple agents collaborating on a shared Canvas investigation, auto-investigation the moment an SLO burn alert fires (preparing visualizations before a human looks), and skills that encode a team's domain expertise so agents and colleagues can lean on them.
Honeycomb is fully distributed, 200+ people, Series D funded, and named to Forbes' America's Best Startups of 2022 and 2023. Customers include HelloFresh, Slack, LaunchDarkly, and Vanguard.
REQUIREMENTS:
- Shipped LLM-based systems people relied on in production (not demos, not fine-tuned models in research context); understand where agent systems break and how to design around it
- End-to-end ownership: can take something from rough prototype to production-grade without handoff
- Current judgment about what's shifted in agent design in the last 6–12 months
- Agent architecture depth: understand how a fast, high-cardinality data store changes what an agent can reason about and how to design for it
- Product judgment: can see what the agent layer does today and envision what it should do next, and make that case with a prototype
NICE-TO-HAVE:
- Observability or developer-tools background (engineers are your users; fluency in that world helps)
- Familiarity with eval frameworks, agent tooling, RAG, and prompt engineering
