Anathem is a VC-backed early stage startup building an AI co-pilot for high stakes services where hours of admin can be saved. Our focus is on transforming services where mental health is the backbone (Psychiatry, Neurodiversity Assessments, Police etc.). Our primary platform currently serves mental health clinicians by automating their admin work — so that they can see more patients and deliver better quality care. Our platform reduces clinician admin work by as much as 75%. With ongoing NHS trials, our mission is to double capacity and clear national waiting lists.
The Role:
Salary expectations: £70k-85k; 0.25-1% equity
We’re looking for a LLM Engineer who can help us develop, architect and optimize the LLM outputs produced for our clinician workflows and our other high-stakes use cases. This covers advanced prompt engineering, prompt development, Retrieval Augmented Generation (RAG), fine-tuning, instruction-tuning to augment LLMs for Anathem.
This role involves staying up to date with the latest research, advancements and techniques for using LLMs and generative AI. You will combine both an empirical scientific approach to validating and evaluating the accuracy of these outputs, with creative problem-solving and strategy to develop better methods in this novel field of LLM prompting/retrieving, together with data engineering best practices to optimize cost and performance.
At Anathem, we will be pioneering the use of generative AI to automate information extraction, retrieval, evaluation and summarization to automate admin work and better utilize data in high-stakes mental health based applications. This includes psychiatry, neurodiversity assessments, police and more. With your expertise in AI research and development, you will help us pave the way towards a further product evolution, eventually one that can enable personalized medicine, predictive diagnostics, clinical decision support etc.
We’re assembling a small, powerful team - so not only will you get the opportunity to work on all parts of the product, you’ll also have a real say on what we build and how we build it. As a Founding Engineer, you will own the ingenuity of your work, be free to creatively problem-solve, and be largely responsible for the creation of a transformative piece of technology that can positively impact the lives of millions.
If that sounds exciting, this is the place for you!
You Will:
Develop, experiment, evaluate and optimize the use of various foundational LLMs such as GPT-4, Claude-3, Mistral etc. for tasks such as question answering, entity extraction, summarization on a range of data sources (transcripts, historic patient records and notes). This includes experimenting and comparing different methods of formulating API queries, prompts, retrievals etc..
Develop strategy and implementation for performing validation of LLM outputs, detecting hallucinations, QA testing for outputs, methods for auditability and explainability, evidences/references to increase trust, legitimacy and accuracy of outputs.
Write code using frameworks such as LangChain, Llama-index, open-source vector databases, Huggingface and other LLM/NLP libraries to apply techniques such as RAG, structured output parsing, function calling, query chaining, templating, vector embedding retrieval etc.
Improve the way we formulate queries, API calls and prompting in order to improve the efficiency of our token usage for both performance and cost effectiveness
Pioneer the development of specialized domain specific LLMs (a bit like MedPaLM2) to enhance accuracy and performance on use-case specific tasks.
Steer the evolution of Anathem’s AI strategy, setting in place LLM validation/testing pipeline to facilitate in our ability to train bespoke AI models for a range of clinicial prediction and decision tasks.
You may be a good fit if you:
Enjoy writing high quality, clean and efficient code
Required: highly proficient in Python, extensive experience in using various LLM/NLP libraries
Desired: experience in Cloud platforms, (e.g. Azure), research experience in ML, familiar with Streamlit
Have 2+ years of Software Engineering experience, and 1+ years of experience with NLP/ LLMs / Machine Learning (preferably in a medical application)
Have developed and architected green field software applications before
Have excellent attention to detail, are meticulous, organized and disciplined with your code development and testing.
Have learned, experimented and developed LLM applications using models such as GPT-4, and frameworks like LangChain and Llama-index
Are passionate and excited about the use of generative AI and LLMs in revolutionizing the world of work, taking pride in developing tools with it that will solve real world problems and drive mass efficiency savings.
A quick learner, independent thinker, and creative problem-solver who keeps up to speed with the latest developments and research on LLMs, and adapt their work appropriately
Thrives in a fast paced, collaborative startup environment that is happy to take on other tasks, where needed, for the greater good of the team.
How to Apply:
Does this role sound like a good fit? Send an email to info@anathem.ai with a CV and your LinkedIn. We look forward to hearing from you!
The interview process will consist of a short interview, a take-home coding task (~3hrs), technical interview, then finally a cultural fit interview before we extend you the offer.