Senior AI and Knowledge Systems Engineer

GeoYeti, a division of Bcore, is hiring a Senior AI and Knowledge Systems Engineer to build an enterprise knowledge platform that connects information across Jira, Confluence, SharePoint, and other sources. The role develops AI-based question-answering workflows, validates outputs with users, and ensures access controls, data protection, and auditability.

Responsibilities:

• Enterprise Knowledge Platform: Design and build a knowledge system that brings together staff reports, plans, and tracked items that today live in separate places across Jira, Confluence, and SharePoint
• Question Answering: Build NLP and LLM-based workflows that let a user ask a plain question, such as “give me a completed trip report” or “show me the burndown from our last trip,” and get an answer written the way the staff would write it
• Staff Validation: Put staff in the loop to review outputs against how they would have answered, and use their feedback to tune the system until they trust it
• Controlled Information Testing: Release only selected information to the model and observe what it surfaces, so we can measure accuracy, find gaps, and understand how the system behaves
• User Testing: Watch users do the task the way they normally would, then test the system’s output against their results and use the differences to guide the next round of work
• Prompt Engineering and Integration: Build and refine prompts, APIs, and MCP connections that link the model to Jira, Confluence, SharePoint, and the other data sources staff already use
• Access Control and Need-to-Know: Make sure the system never shows a user anything they aren’t cleared or authorized to see. Answers must respect the permissions already set in Jira, Confluence, and SharePoint, along with classification markings, compartments, and need-to-know. This applies to every part of the system, including search indexes, embeddings, prompts, and model outputs. You will also build in audit logging and keep the data on the networks where it belongs
• Working With Real-World Systems: Work with the systems as they are, not as the documentation says they should be. Some installations are not on current versions, APIs may be limited or unreliable, and MCP connectors may not exist for what you need. You will find workable paths, such as building your own connectors, adapting to older APIs, or pulling content another way, and document what you learn
• Path to Enterprise Capability: Grow the first working version into an enterprise knowledge store, working toward initial and then full operational capability