Job Description:

The Role

You are a seasoned AI engineer with a deep understanding of how LLMs work and can be integrated into a modern information-processing stack.

You Will Have

• Significant, demonstrated experience in solving complex business tasks by harnessing state-of-the-art LLMs to extract, analyze, summarize, and otherwise operate with information in a large variety of document types.

• Excellent knowledge of Gen AI tool stack — understanding how to parse and chunk documents for specific kinds of application tasks, embeddings, retrieval augmented generation, designing complex information-processing flows with LLMs inside, crafting text instructions to reliably elicit desired task behavior from LLMs

• Deep expertise in Python, in particular being productive with idiomatic Pythonic techniques and standard libraries, so that sophisticated systems can be built quickly with a small amount of code.

• Excellent knowledge of standard tools of the trade for machine learning engineers, such as WandB for ML flow, Prodigy, and open-source tools for data/document annotation.

• The ability to design, run and evaluate a series of experiments that will enhance your team's understanding of how specific LLMs can be used to accomplish specific tasks.

• Bachelor’s in computer science or related area, plus five years of experience, or a Master's with two years of experience.

• Experience in finance, broadly, and investing specifically is highly desirable.

The Team

We are developing a Large Language Model (LLM)-first knowledge stack for investment professionals in Fidelity Asset Management – analysts and portfolio manager in equities, fixed income, high income, direct lending. The stack will be able to process all the documents of interest to analysts – e.g., analyst reports, earnings notes, spreadsheet models, prospectuses, loan indentures, news reports, regulatory filings. It will support the deployment of personalized assistants that can assist principals in their full range of information consumption/processing/production tasks.

These tasks include summarizing documents (from different points of view), answering questions about passages, extracting terms from documents, comparing extracted information across time, across deals, consolidating and restating agreements (with their amendments), identifying trends, monitoring changes in the state of the world that are relevant to a particular professional, generating plausible scenarios, analyzing and decomposing into factors, generating reports. Assistants are personalized to their principal, deliver results with high quality, and are expected to get better over time, with increased interaction with their principals and with the world.

Certifications: