Principal AI Engineer - Context - Agents and Context
Principal AI engineer owning the end-to-end production improvement loop for Elastic's Context Engine — the knowledge layer that AI agents use to work with enterprise data in Elasticsearch. Ships TypeScript production code and designs the evaluations, telemetry and safe rollout practices for agents, prompts, skills and automations.
- Compensation
- Not disclosed
- Location
- United Kingdom
- Work setup
- Hybrid
- Seniority
- Principal
- Employment
- Not specified
- Experience
- 10+ years
- Category
- Ai engineering
- Posted
Skills
Must have
- 10+ years of software engineering experience, recent years shipping and operating AI-driven products on real production traffic
- Track record of eval-driven product improvement: diagnosing agent/LLM behaviour from traces, designing evaluations, shipping fixes and measuring outcomes
- Direct experience building agents with state and memory and iterating on prompts, skills and tool behaviour safely in production
- Familiarity with MCP, including exposing public MCP servers and tools
- Experience designing telemetry for AI systems and using it for engineering and product decisions
- Experience running product experiments end to end: instrumentation, unattended execution, interpreting results
- Experience building products with public APIs and evolving data models without breaking customers
- Strong backend engineering skills in Python or TypeScript: APIs, stateful workflows, data pipelines, production services
- Comfort working with data scientists, engineers and product managers as peers, communicating trade-offs clearly
- Pragmatic, low-ego style suited to a distributed, async-first team
Nice to have
- Experience with agent frameworks such as LangGraph, CrewAI, Claude Agent SDK, or similar
- Agentic retrieval experience, including knowledge representation
- Practical Elasticsearch experience
Full description
Elastic, the Search AI Company, enables everyone to find the answers they need in real time, using all their data, at scale — unleashing the potential of businesses and people. The Elastic Search AI Platform, used by more than 50% of the Fortune 500, brings together the precision of search and the intelligence of AI to enable everyone to accelerate the results that matter. By taking advantage of all structured and unstructured data — securing and protecting private information more effectively — Elastic’s complete, cloud-based solutions for search, security, and observability help organizations deliver on the promise of AI.
What is The Role
The Context Engine team builds the knowledge layer that AI agents use to work with enterprise data in Elasticsearch. We extract knowledge from any data sources into a structured AI Index, serve it to agents through public APIs, MCP tools and framework integrations, and close the loop with agent traces so that what the engine knows improves from real usage. Any agent can use it: Elastic’s own Agent Builder, Claude Code, LangChain and other third-party harnesses.
As a Principal AI Engineer, you own the improvement loop of this product end to end: how agents, automations and skills behave in production, how we observe them, how we evaluate them, and how we ship changes to them safely. This is a hybrid role at the intersection of engineering, data science, and product. You will write production code, design evaluation and telemetry that product decisions can rest on, and set the technical bar for how the team iterates on agentic behaviour. You will work alongside data scientists, backend engineers, product, and UX, and your work will show up directly in what customers build on top of Elastic.
The codebase is TypeScript and we build it in the open, so you'll be shipping code, designs and discussions in public alongside the rest of the Elastic Stack.
What You Will Be Doing
- Own the production improvement loop for Context Engine: understand how extraction automations, retrieval tools and memory behave, based on offline evaluations and customer conversations and telemetry. You help find the failure modes, fix them, and prove the fix.
- Define how we iterate on agents and skills safely: versioning and rollout of prompts, skills and automations, regression coverage, staged and shadow evaluation, and the guardrails that let us change behaviour without breaking customers.
- Design the telemetry we need to make data-informed engineering decisions: what to capture from agent traces, tool calls and knowledge retrieval, how it lands in Elasticsearch, and how it feeds evaluation, dashboards and the feedback loop.
- Partner with the data science team on evaluation strategy: golden datasets, evaluators to gate on quality, latency and cost.