ai engineer

The company’s American fitness equipment partner is developing a next-generation application platform designed as an AI-driven, highly personalized global training platform integrated with wearables, medical data, and fitness devices.

• Optimize latency and cost through content pre-generation, session-revision-keyed caching, and content-fingerprint deduplication;
• Maintain and expand evaluation suites, including model-based judging and adding test cases after each production incident;
• Implement the personalization pipeline, including per-turn signal extraction, profile synthesis, and epoch-based cache invalidation;
• Write and maintain FastAPI services, Celery workers, Pydantic schemas, and pytest tests;
• Design and develop agent graph topology and tool contracts in LangGraph/LangChain;
• Build and tune retrieval using embeddings, hybrid ranking, and hard pre-filters over a vector store;
• Monitor production traces, token usage, quality metrics, and feature-flagged rollouts.

• 5+ Years of production software development experience;
• Senior-level Production Python 3.11 experience with typed async services;
• Experience designing LLM agents in LangGraph/LangChain or a comparable framework, including graph topology and tool contracts;
• At least 3 months of production responsibility for an AI/ML/LLM function, including agent systems and tool-calling;
• Ability to work independently and take ownership in an environment without ready-made specifications;
• Full ownership of FastAPI services, including endpoints, workers, schemas, and tests;
• Experience with RAG and retrieval, including embeddings, vector search, and hybrid ranking;
• Experience with personalization and ranking based on behavioral signals;
• English communication skills;
• Nice to have: LangChain, LangGraph, Celery, AWS, MongoDB Atlas Vector Search, OpenSearch, Redis, Postgres, Docker/Terraform, Datadog/Opik/Amplitude, TypeScript/NestJS/React Native, bandits/uplift modeling, voice products/TTS.

• Project-based cooperation with payment based on stages and acceptance protocols for individual deliverables;
• Administrative and settlement support from Connectis;
• Access to Claude Enterprise with unlimited tokens, GitHub Enterprise, and full AWS infrastructure;
• Participation in industry events and technology meetups;
• Flexible B2B cooperation;
• The application process includes an AI Recruiter available 24/7; the final decision is made by the Project Manager.