• Development of LLM components for procurement processes and OCR/data extraction pipelines from documents.
• Building RAG search for contracts and specifications using embeddings, rerank, and chunking.
• Implementation of structured output (JSON Schema/pydantic), model response validation, hallucination reduction.
• Rapid prototyping of open-source models (Hugging Face/GitHub) and creating demos on Gradio/Streamlit.
• Integration of ML services via FastAPI API, CI/CD setup, logging, and quality metrics.

Required:

• Python, PyTorch, Hugging Face Transformers, CUDA/GPU inference.
• REST API, Docker, configurations, secrets; experience with queues - a plus.
• OpenSearch / Elasticsearch / PostgreSQL.
• Ability to measure the quality of ML/LLM components.

Bonus:

• VQA, vision-language models, multimodality.
• Experience with Stable Diffusion / SDXL / ControlNet / LoRA is welcome.

A test assignment is required!!!