NativeCall 10 дней назад
AI/Python-разработчик
- Москва
- Удалённо
PythonAIDockerCI/CDGitLinuxLLMOpenAI APIHugging FaceAnthropic
Описание
Responsibilities
- Design and develop LLM agents and multi-agent systems in conjunction with the development team.
- Write clean, structured, and maintainable code in accordance with standards and best practices.
- Create and maintain documentation for code, processes, and applications.
- Track the development of web technologies and industry trends, propose new solutions.
- Optimize web applications for speed, scalability, and security.
- Participate in the full development lifecycle – from concept to deployment and maintenance.
Requirements
- Python 3.10+ at a confident developer level.
- Experience with LLM APIs: OpenAI, Anthropic, open-source models via vLLM/TGI/Ollama.
- Practical experience in model fine-tuning: SFT, LoRA/QLoRA, DPO/ORPO; preparation and validation of training datasets; quality assessment, combating overfitting and degradation of basic skills.
- Experience with HuggingFace Transformers / PEFT / TRL.
- Building evaluation systems (evals): designing task-specific metrics, regression test sets, LLM-as-judge with an understanding of its limitations, A/B testing.
- RAG pipelines: chunking strategies, hybrid search, reranking, vector databases (pgvector, Qdrant, Weaviate or similar); understanding when RAG is preferable to fine-tuning and vice versa.
- Orchestration of data and training pipelines: Airflow, Prefect, Temporal, or similar.
- Observability of LLM systems: tracing and monitoring (Langfuse, LangSmith, OpenTelemetry, or similar).
- Docker, Git, CI/CD, Linux; basic understanding of working with GPUs (VRAM management, quantization, inference servers).
- Fundamentals of LLM application security: protection against prompt injection, tool isolation, principle of minimum privileges for agents.
Nice to have
We offer
Hiring process
- Submit your resume.
- We will review it and contact you.