We are hiring an AI engineer — Sofia or Israel.
We have AI agents doing real work against production codebases today: they take a work item, plan it, write the code, review it, and open a pull request for a human to approve. What that system does not have is anyone whose job is to make it measurable.
Responsibilities:
Building the agents — new ones, and extending the ones already running.
Their tools, and the context architecture that decides how each agent behaves.
The evaluation harness that catches an agent getting worse before the change merges.
The telemetry that proves what any of it is actually worth.
Budgets that govern model spend instead of observing it.
The retrieval layer that lets an agent find what it needs instead of reading an entire repository to answer one question.
The runtime underneath is moving from self-managed Kubernetes to Bedrock AgentCore, you will own the second half of that migration.
Requirements/Tools:
Experience with the following:
Python – used across the service and agent codebases.
Temporal – used for orchestration — multi-stage pipelines that run for hours with human approval gates in the middle.
LangGraph and Claude Code (agent runtimes).
AWS Bedrock – for models behind a LiteLLM gateway that keeps provider choice open.
Langfuse – used for LLM traces and evaluation.
OpenTelemetry with Prometheus and Grafana for everything else.
PostgreSQL with pgvector used for memory and caching.
MCP servers for tools.
Azure DevOps Repos and Pipelines for delivery.
Experience in startup environments would be considered an advantage
Experience with the following:
Python – used across the service and agent codebases.
Temporal – used for orchestration — multi-stage pipelines that run for hours with human approval gates in the middle.
LangGraph and Claude Code (agent runtimes).
AWS Bedrock – for models behind a LiteLLM gateway that keeps provider choice open.
Langfuse – used for LLM traces and evaluation.
OpenTelemetry with Prometheus and Grafana for everything else.
PostgreSQL with pgvector used for memory and caching.
MCP servers for tools.
Azure DevOps Repos and Pipelines for delivery.
Experience in startup environments would be considered an advantage
