Research Engineer: ML Deployments & LLMs (Hybrid)

Please see our Company Principles to understand the core things we value – in particular, we are looking for exceptional people who are willing to tackle some of the most difficult technical problems there are, in order to create something extraordinary with huge impact.

As a Research Engineer at Unlikely AI, you’ll assist in delivering model prototypes to production. You’ll play a key role in product delivery and designing and implementing new ML features on our platform, which typically includes managing model deployments and ensuring stability. Other projects could include optimising our vector search capabilities.

You should have a core understanding of ML fundamentals, and be up to date with the latest LLM models to undertake evaluation of new implementations. As a growing startup, this role could include projects beyond the scope of this job description therefore we are looking for individuals who are versatile and enthusiastic about learning new skills as our Applied Science team evolves.

This role includes:

• Implementing, deploying, and monitoring deep learning models, including LLMs.
• Optimising model deployments and designing deep learning model features systems.
• Conducting comprehensive performance evaluations, focusing on latency and accuracy across different implementations.
• Communicating complex solutions to colleagues, facilitating collaboration and knowledge sharing.
• Analysing and inspecting large-scale datasets, effectively managing data scalability and integrity.

Required:

• Experience utilising & deploying deep learning models.
• Strong coding skills in Python, including the use of PyTorch or TensorFlow.
• Enthusiasm to learn and get up to speed with cutting-edge technologies that you may not already be deeply familiar with.
• Strong verbal and written communication skills.
• Experience with cloud infrastructure (e.g. AWS / GCP / Azure).
• Experience with MLOps, with strong expertise in Docker for containerization and orchestration.
• Knowledge of ML model deployment including technologies such as Torchserve, Sagemaker or VertexAI.
• Understanding of modern best practices for agile software development.
• Knowledge of the latest developments in NLP including LLMs and the transformer architecture.
• SRE: An understanding of how to keep models stable and performant in production settings.

Desirable:

• Experience with building CI/CD workflows.
• Experience working in a startup.
• Experience with retrieval augmented generation for LLMs and semantic vector search.
• Experience optimising model deployments in terms of latency and throughput.
• Infrastructure-as-code tools, such as Terraform.

Please note this role is not a pure research role and does not involve the creation of academic literature, but you should be very comfortable with reading and utilising academic papers and applying these concepts in your work.

Location: We are currently operating a hybrid scheme with a small office near Holborn tube station available to anyone who wants to work there. We also have occasional team days where everyone meets face to face and days where people work heads down from home, communicating with colleagues using Slack and Zoom.

Compensation: Compensation will be through salary and generous share options. The company has a tax-efficient EMI share option scheme set up (not available to larger companies) which allows us to provide real exposure to the success of the company without taxes being due when they are paid.