Senior AI Engineer, Agents

This is a fully remote position, open to applicants in Greece, +4 more countries.

📋 Description

• Develop and sustain the agent roadmap by pinpointing repetitive expert tasks in collaboration with sales, data, and product teams.

• Evaluate potential agents based on feasibility, cost, and impact.

• Architect agent designs and create production code for tools, loops, sub-agents, memory, state management, and evaluations.

• Construct a reusable platform layer over EveryWatch data, encompassing pricing, auctions, listings, references, and portfolios.

• Create an MCP interface atop the current backend.

• Establish shared memory, tracing mechanisms, and reusable evaluation infrastructure.

• Enhance WatchChat's multi-turn state, memory management, latency, cost-efficiency, tool-call reliability, and regression checks.

• Generate high-quality multi-turn datasets, programmatic tool/argument verifiers, and LLM-as-judge evaluations on held-out datasets.

• Streamline model selection, context budgets, caching strategies, and fine-tuning to effectively manage costs and latency.

• Collaborate with engineering, product, sales, and data teams to identify and implement AI capabilities for collectors, dealers, sales teams, data operations, and engineering.

⛳️ Requirements

• Over 6 years of experience delivering production software, including at least 2 years working with LLM systems utilized by actual users.

• Proficient in ReAct or similar loops, tool/function invocation, planning, state and checkpointing, long-term memory, HITL processes, and streaming.

• Practical experience with LangGraph or a compelling case for an alternative solution.

• Familiarity with tracing and observability stacks such as LangSmith or Langfuse.

• Strong background in production Python, including FastAPI, asynchronous job management, and clear service boundaries.

• Solid experience with proper RAG: retrieval, re-ranking, relevance assessment, and evaluation.

• Knowledge of RAGAS, DeepEval, GEVAL, LLM-as-judge, or pass@k evaluation methodologies.

• Cloud production experience with AWS (Bedrock, SQS, EC2/EKS, S3) or a comparable platform.

• Experience with Docker and CI/CD processes.

• Strong architectural insight and the ability to question existing technical choices.

• Familiarity with LLM post-training methods such as SFT, LoRA/QLoRA, preference/RL techniques, reward design, or verifier design is a plus.

• Experience in voice agents, vision-language projects, multi-agent coordination, MCP integrations, text-to-SQL, report/document generation agents, guardrails, PII management, hallucination detection, or risk scoring is advantageous.

• A passion for watches, collectibles, or market data is a bonus, but not mandatory.

🏝️ Benefits

• Full ownership of EverWatch's entire agent layer and the roadmap associated with it.

• Access to a unique dataset that is unavailable elsewhere.

• Direct communication with the CTO; decisions made in days rather than quarters.

• Budget allocated for models, tools, and the coding agents you wish to utilize.

• Competitive salary package.

• Remote-first work arrangement.