Sabrine Krichen

Collaborated with a team to solve the 3D bin packing optimization problem using reinforcement learning, developing an environment that handles complex constraints and requirements.

Developed maintainable and reusable Python code for production-level RL pipelines, improving product performance by 10%.

Played a key role in transitioning the product from research to production.

Designed a prompt-driven, multi-agent protein design assistant using GPT models, ProteinMPNN, LangChain/LangGraph, and structured role-based prompts with constraints, error-handling, and inter-agent communication protocols.

Demonstrated strong prompt and context engineering across top-tier LLMs, enabling reliable tool-calling, long-context handling, and multimodal workflows involving structural (PDB) data.

Integrated MCP server and tool-calling capabilities to let agents fetch structured protein data, run external bioinformatics tools, and interact with scientific APIs programmatically.

Built a production-grade RAG pipeline powered by Weaviate, leveraging optimized embeddings and high-performance vector retrieval to deliver more accurate, scientifically consistent LLM responses.

Collaborated with BioNTech and multidisciplinary teams to leverage generative AI and deep learning for protein design and nanoparticle optimization, advancing vaccine development.

Developed generative AI models for protein structure inpainting and enhanced protein-based nanoparticle designs using diffusion models and sequence-to-sequence transformers.

Integrated PostgreSQL to store nanoparticle designs, oligomer sequences, docking scores, and AlphaFold2 predictions, enabling efficient querying, experiment tracking, and large-scale dataset management.