Senior Staff AI Engineer Prague, Czechia
Productboard · Prague
Senior Staff AI EngineerPrague, CzechiaLocation: Prague, Czechia (hybrid). Brno or remote within the EU is possible.
Relocation support is available. Only the best engineers are being built here. At Productboard, we are defining AI-first Product Engineering.
Not as a one-time achievement, but as a baseline we raise every time we reach it. We remove every bottleneck as it surfaces and rebuild how product teams work from first principles, not just adding AI to existing workflows.AI tools are everywhere now. Agents can write code.
That’s today’s baseline. What we're building is the environment where the baseline makes you better: focused on outcomes, writing specs that agents build from, leveling up with every feature you ship. We are defining what AI-native product development becomes.
We're all in. What you will buildYou'd lead the architectural work that makes our codebase AI-native. Specifically:Agent-native architecture standards.
Clear API contracts, semantic naming, well-defined module boundaries. The patterns that keep AI effective as our systems grow.A context infrastructure layer. Repo-versioned guidance that AI tools load automatically — improving the output of Cursor, Claude Code, and Codex simultaneously.
Self-healing workflows for incidents. Agents that triage alerts, pull logs, surface relevant history, suggest remediation. Humans intervene where judgment matters.
Agent workflows that go beyond chat. Multi-step reasoning, tool use, autonomous task execution with human-gated checkpoints — for product features, not internal demos. Systematic optimization of AI code review.
Earlier detection of correctness, security, and maintainability issues, with the noise filtered out before it reaches a human reviewer. The problemProduct managers (PMs) get more customer feedback than they can read. Today AI processes only part of it.
Insights can arrive weeks after the customer spoke, and PMs still link each need to strategy by hand. We want AI to process all customer feedback and tell each PM what customers need, why it matters, and how it connects to strategy. When it works, a PM opens Productboard and sees: "'Slow dashboard loading' emerged across 23 customers (USD 1.4M ARR).
Maps to Objective 'Improve Performance'. Competitor X shipped performance improvements last week." The PM accepts the insight and asks Spark, our AI agent for product teams, to draft a spec. You will own the AI architecture that produces that insight, and set the AI patterns that other domains at Productboard follow.
The outcomes we measureFeedback informs decisions automatically. Themes map to objectives, initiatives, and the product hierarchy. PMs accept, reject, or refine them.PMs prioritise by business impact.
Each theme shows annual recurring revenue (ARR), open deals at risk, and links to customer churn. Competitor moves show up next to feedback. When a competitor ships something, PMs see which feedback themes it touches.
Insights makes Spark smarter. Spark's answers are only as good as Insights data. Targets:
PMs accept more than 80% of AI insights, more than 90% of feedback connects to a feature, objective, or initiative, and nobody processes feedback by hand. We have not reached these numbers yet. Reaching them is the job.
What makes this hardChanges travel. Insights data feeds Spark, so a change in how we group feedback changes what Spark tells PMs. You weigh those effects across domains before a change ships.
Messy ground truth. Often there is no single right answer to "which theme does this note belong to?" Two good PMs may disagree. You decide how we measure quality honestly.
Cost. Every AI call has a price, and customers pay for AI usage. As usage grows, you decide where a smaller model, caching, or skipping work keeps quality and protects margin.
Influence without authority. Most teams you guide do not report to you, and some are new to building with AI. You lead through clear arguments and clear documents.
What you will doOwn the Insights AI architecture and its technical direction: what context each step gets, which tools agents may call, and how we know an output is good. Set the quality bar for AI in Insights, and make evals the gate every change passes before it ships. Lead AI architecture decisions that span Insights, Spark, and the platform.
Bring teams with different priorities to one written decision, such as the agent conventions every domain follows. Drive the largest Insights initiatives as the named driver. Take a new AI feature to customer beta: set the milestones and acceptance criteria, agree scope with the PM, and decide what ships.
Write production code on the hardest parts, every week. Grow the engineers around you. Your docs and reviews let any engineer pick up an AI workflow and change it safely, without waiting for you.
How we workProductboard runs AI-first product engineering. Coding agents write much of the code; engineers own design, review, and results. Each initiative has one named driver and starts from a one-page pre-read.
We ship to internal users first, then beta, then everyone. We commit the roadmap about a month ahead, and PMs and designers ship to production too.AI layer: Python, Pydantic AI, Vespa, Braintrust, Temporal, LaunchDarkly, Datadog, and Anthropic models.
Around it: Kotlin (some legacy Ruby), TypeScript, React, GraphQL, PostgreSQL, MongoDB, Elastic, Redis, Snowflake, AWS, Kubernetes, and Terraform.
What you need
Typically 10+ years building production systems, with strong backend skillsYou have designed LLM systems that customers use, measured their quality with evals, and made the trade-offs between quality, latency, cost, and reliability. Hands-on experience with search, retrieval, clustering, or entity matching over large text collections. Deep experience with distributed systems and long-running workflows.A record of setting technical direction across several teams or domains, including teams that did not report to you.
Clear written English. You can explain a technical decision on one