Job Summary

The goal of this task is to design and build two chatbots that utilise custom data. One to allow an operator to rapidly access or obtain a summary of tactical information. Another will aid developers and suppliers in using the Faster platform for application development and task delivery.

Machine learning (ML) algorithms have revolutionised many industries in recent years, notably using computer vision and large-language models. Recently, a technique called Retrieval-Augmented Generation (RAG) has been used to enhance large language models and appears an ideal solution to improving chatbots for custom data.

Two candidate application of RAG are proposed:

1) Operator Queries: In the underwater space, sonar and tactical operators have limited time to make informed tactical decisions as to the detection and classification of friendly and threat platforms based on the information available to them. This information can come from a number of different places but include data catalogues such as the "Janes' Fighting Ships" publication which provides identification profiles for various maritime vessels and aircraft. There are also RN specific publications that would provide a custom set of data for ingestion into the RAG pipeline.

2) Developer Queries: Separate to this use case, the Faster platform is built of a number of components, each with accompanying documentation as well as guides on how to develop applications and troubleshoot developing on the platform. A chatbot using the RAG system would be used to improve developer productivity; to make access and queries of this documentation as efficient as possible for developers and suppliers.

Responsibilities & Duties

The supplier should carry out an initial investigation into the current state of RAG pipelines with LLM's in open-source repositories and literature.

Supplier developers will be on-boarded to the Faster Platform and have access to a KASM remote desktop environment with suitable development tools like VSCode. A suitable set of data will be provided to the supplier by Faster for each of the chatbots.

The supplier is to create pipelines for each of the chatbots, including the ingestion of the provided data and suitable chunking algorithm for the data. A series of different experiments should be run utilising different LLMs and consideration should be given to how the chatbot answers are evaluated, e.g. evaluation of the initial retrieval as well as the whole system.

A simple UI should also be developed for an operator or developer to send queries to and display the response.

The chatbots created in this work will be standalone applications, deployed to the FASTER Platform, with an associated API for users to call.

Essential Skills

·         MSc, or PhD in Speech & Signal Processing, Computational Linguistics, Computer Science, Data Science, or related fields.

·         Proven experience in AI/ML development, with a focus on ChatBot technologies. Experience in the public sector is a plus.

·         Proficiency in programming languages such as Python, R, or Java. Strong knowledge of NLP, machine learning frameworks (e.g., TensorFlow, PyTorch), and cloud platforms (e.g., AWS, Azure).

·         Ability to analyze complex data sets and derive actionable insights to improve ChatBot performance.

·         Excellent verbal and written communication skills, with the ability to explain technical concepts to non-technical stakeholders.

·         Strong problem-solving skills and the ability to work independently and as part of a team.

Nice to have Skills

·         Prior working experience in defence industry.

Background Check Required

BPSS minimum security clearance

Project Budget

·          Budget – Day Rate: £ 525.00/Day plus VAT (Milestone based Payment)

·          Expenses Allowed: No

·          Extension: No

·          Language: Fluent in English.

·          Outside of IR35 regulation.