Build the agents at the core of a corporate learning platform — planning, tool use, memory — portable across Claude, OpenAI and Gemini.

• Design and ship agents — the core of the role: planning, tool use, memory, and guardrails, orchestrated with LangChain/LangGraph, exposed via MCP where it fits, running behind a provider-abstraction layer with model routing and cost control
• Build product features end to end — from the front end through backend services to the model call in the cloud
• Set up and tune retrieval: ingestion, chunking, retrieval quality, reranking, and grounded generation with citations
• Evolve the content-generation pipeline: grounded generation from ingested sources, structured outputs that survive provider differences, and accuracy checks that keep generated material true to source
• Develop hands-on lab environments in a sandboxed execution platform: mock APIs, auto-verification harnesses that grade learner work, and managed multi-provider model access with per-user budgets
• Take charge of evaluation infrastructure: task datasets, deterministic and model-based graders, regression suites, and cross-provider benchmarks

• 4+ years in software engineering, with production systems you can walk through end to end
• 1+ years shipping production LLM applications — agents, tool use, retrieval — with hands-on work across at least two of the three major platforms (Anthropic, OpenAI, Google)
• Depth in agent development: tool use, memory, multi-step orchestration, and MCP — you've built and debugged MCP servers, not just consumed them
• Claude Code as a daily working tool, plus working fluency with the OpenAI API and Gemini — or a demonstrated ability to get there fast, since the abstractions matter more than any single SDK
• Evaluation fluency: task datasets, deterministic and model-based graders, regression suites. "I tested it manually and it looked fine" is an unfinished sentence
• Solid Python for LLM tooling, with experience in LangChain/LangGraph or an equivalent orchestration framework
• Comfort with sandboxed cloud execution environments, CI, and API security basics
• Clear technical communication — you can review someone else's work rigorously and kindly, and explain a model limitation to a non-engineer without jargon
• A responsible, outcomes-focused mindset
• Advanced English or higher

Recruitment Process:

• HR&Technical Interview
• Client stage
• Offer

• Technical Ownership: You own the AI architecture and the standards behind it
• Modern AI Work: Agents, retrieval and evaluation as the everyday job, not a side experiment
• Collaborative Environment: A team that values partnership, creativity, and mutual respect
• Flexible Work: Work remotely from the comfort of your home or join us in our modern Kyiv office
• Generous Time Off: 20 paid vacation days + 15 sick leave days annually
• Professional Growth: Compensation for courses, certifications, and learning resources
• Cutting-Edge Tools: Access to premium AI tools (Cursor Pro, Claude Code, GitHub Copilot, etc.)