#remote #vacancy
Company: Excdev
Position: ML Engineer
Salary: 2500-3800 USD
Project Stack
Backend:
Python, FastAPI, SQLAlchemy, asyncpg, PostgreSQL, Alembic
ML / AI:
GUI-OWL (UI-TARS), GPT-5-mini, Claude (Anthropic Computer Use API), vLLM, OpenAI-compatible APIs
Agents:
proprietary CUA pipelines (GUI-Owl, UI-TARS agent loop, Anthropic computer-use)
Infrastructure:
Docker, Docker Compose, S3 (logs and screenshots), VNC, virtual machine management via Docker API
Responsibilities:
- Develop and optimize CUA agents: increase scenario completion accuracy, reduce the number of steps, improve handling of edge cases (captchas, non-standard UI, dynamic content).
- Design and implement new agent pipelines (multi-agent, judge-based architectures).
- Work with vision-language models (UI-TARS, Claude Vision): selection, fine-tuning, prompt engineering, quality assessment.
- Integrate and deploy LLM services (vLLM, OpenAI API, Anthropic API), optimize inference (tensor parallelism, batching).
- Participate in system scaling: increasing the number of concurrently processed tasks, task parallelization, virtual machine resource management.
- Work with data: parsing, structuring results, integration with PostgreSQL and S3.
Requirements:
Mandatory:
- Experience working with LLMs in a production environment (prompt engineering, function calling, structured output).
- Experience building AI agents (LangChain / LangGraph / ReAct / Tools).
- Understanding of CUA / GUI-agent architecture and principles (Anthropic Computer Use, UI-TARS or similar).
- Proficient Python skills (asyncio, FastAPI or similar frameworks).
- Experience with Docker (building images, docker-compose, container management).
- Ability to read and reproduce ML research results (papers, benchmarks, open-source models).
Will be a plus:
- Experience deploying and optimizing LLM inference (vLLM, TGI, tensor parallelism).
- Experience fine-tuning vision-language models.
- Familiarity with multi-agent systems and agent orchestration.
- Experience working with the Anthropic API (including Computer Use).
- Understanding of web automation (Selenium, Playwright, pyautogui).
- Experience with PostgreSQL, SQLAlchemy, Alembic.
Conditions:
- Work on an R&D project in the field of AI agents and LLM systems.
- Modern tech stack: Python, LLM services, vision-language models, agent architectures.
- Opportunity to work with cutting-edge solutions in Computer Use Agents.
- Remote work format from any city or country.
- Work schedule: 5/2.
- Vacation: 28 calendar days.
