Data Scientist (LLM)

What You'll Do:

• Fine-tune and deploy Large Language Models (LLMs), including Retrieval-Augmented Generation (RAG), to develop cutting-edge workflow tools that directly impact millions of customers every day.
• Be an integral part of a dynamic, startup-like team within our R&D department, thriving in an environment of chaos, camaraderie, and creativity.
• Continuously absorb and apply the latest in NLP, LLMs, and RAG, driving innovation and excellence in our solutions.
• Experiment, research, and benchmark to refine model architectures, optimizing performance and efficiency with a specific focus on the integration of retrieval systems in LLMs.
• Technologies you will work with on a daily basis include: Hugging Face's libraries (such as Transformers, Accelerate, Datasets, and PEFT) and CLI, PyTorch's torchrun, GCP for Cloud Computing, Kubernetes for scalability, Tim Dettmers' bitsandbytes, standard data analysis tools (Pandas, NumPy, SciKit-Learn, Matplotlib), Git, vector databases (Qdrant, Pinecone), and Bash scripting.
• Bonus: we love Weights and Biases and MLflow.

Key Responsibilities:

• Innovate and improve LLM training and serving methodologies, with a special emphasis on integrating retrieval mechanisms in the models.
• Stay relentlessly updated with industry trends, new techniques, and novel applications of LLMs, particularly in the context of RAG.
• Develop, test, and deploy LLMs, including those with retrieval-augmented capabilities, for various generation tasks, ensuring seamless integration and high performance.
• Share your knowledge and findings on RAG and other LLM advancements, fostering a culture of learning and growth within the team.
• Tackle challenges head-on during model training and deployment, including the unique aspects of integrating retrieval systems with generative models, ensuring smooth and effective resolutions.

What We Expect From You:

• While we are looking for someone with a strong foundation in LLMs and RAG, we understand that the field of AI is vast and constantly evolving. We value candidates who have a solid base in some of the key areas and are enthusiastic about learning and growing in others.
• Foundational Knowledge: A strong understanding of machine learning, NLP, and basic principles of neural networks. Familiarity with the concepts behind large language models and retrieval-augmented generation is a plus.
• Technical Experience: Hands-on experience with some of the technologies we use, such as Hugging Face’s libraries, PyTorch, GCP, Kubernetes, and data analysis tools. If you’ve worked with vector databases or have experience in Bash scripting, that’s a bonus.
• Problem-Solving Skills: The ability to tackle complex challenges, think strategically about resource management, and optimize model performance.
• Learning Attitude: A keen interest in staying updated with the latest trends and developments in AI and the willingness to continuously learn and apply new techniques and methodologies.
• Team Collaboration: Strong communication skills and the ability to work collaboratively within a diverse team, sharing knowledge and contributing to a learning culture.