Artificial Intelligence Researcher

About the Position

A dynamic FinTech startup seeks an Artificial Intelligence Researcher and Machine Learning Engineer to develop, retrain, and enhance large language models (LLMs) aimed at transforming intricate financial markets. The role offers close collaboration with the founders and founding engineers, providing significant autonomy in problem-solving approaches.

Key Responsibilities

• Designing, developing, and deploying LLM-driven solutions using PyTorch.
• Fine-tuning open-source models utilizing techniques such as LoRA, QLoRA, and PEFT.
• Creating and refining embedding and reranking models.
• Constructing datasets and generating synthetic data for training purposes.
• Building semantic search and document intelligence systems.
• Conducting experiments and implementing comprehensive evaluation frameworks for models.
• Optimizing model performance for real-world production environments.
• Collaborating closely with domain specialists to translate complex workflows into machine learning models.

Required Skills and Experience

• Proven hands-on experience with PyTorch and transformer-based models, including BERT, RoBERTa, and Llama.
• Expertise in fine-tuning LLMs using LoRA, QLoRA, or PEFT.
• Working knowledge of embedding and reranking (cross-encoder) models.
• Experience with semantic search and information retrieval systems.
• Competency in designing and cleaning datasets and generating synthetic training data.
• Strong background in conducting ML experiments and evaluations.
• Familiarity with Hugging Face, Transformers library, or comparable tools.
• Understanding of production-level ML deployment including containers and cloud infrastructure.

Additional Details

This role is not focused on building AI wrappers or prompt engineering. Instead, the ideal candidate is someone who has thoroughly worked with building, tuning, and deploying ML models, relishing the challenge of complex problem-solving in novel domains.

Academic, research, or R&D backgrounds are definitely valued. While experience with Retrieval-Augmented Generation (RAG) is helpful, the position is not intended for candidates whose main experience is connecting APIs with frameworks such as LangChain.

The company prioritizes candidates who understand the underlying mechanics of the RAG stack, including embeddings training and evaluation, retrieval and reranking operations, dataset preparation, and model improvement through rigorous experimentation.

The position requires presence in London four days a week in a hybrid work model and offers a compensation package up to £100,000 plus potential bonuses.