Eddie Simmens

AI is powerful. Making it useful takes more than code.

I help organizations identify where AI can create value and turn promising opportunities into working systems and workflows. I bring hands-on experience building AI agents and machine learning models, along with an understanding of people and how work gets done, to determine what to automate, where to augment human judgment, and how to evaluate the results.

See my work

Featured Projects

01 / 03 View project

Slack AI Deal Approval Workflow diagram in three stages: Capture (AI conversation), Route (deterministic rules), and Approve (human decision), ending in an audit log and a Slack notification.

AI Agent for Sales Approvals Deal Approval Workflow, 2026

Built an AI agent for Slack that guides sales reps through requesting approval for deals that fall outside company policies. The AI gathers the deal information conversationally, while a separate Python service determines who must approve the deal. In automated test conversations, the system correctly identified all required approvers 29 out of 30 times, without missing any mandatory approvals.

AI-Powered Strep Throat Screening Computer Vision, 2026

Built a deep learning system that analyzes smartphone throat images to help screen for strep throat. Developed in collaboration with three physicians, it was designed to flag uncertain predictions for a doctor to review rather than relying on AI alone. The pipeline outperformed all 20 previously published machine learning baseline models.

Where Generative AI Fits at Work Workforce AI Research, 2026

Conducted independent research to understand where generative AI is best suited to different types of work. Analyzed data derived from 200,000 Microsoft Copilot conversations and U.S. workforce data covering 738 occupations, using machine learning to predict which types of work are better suited to generative AI. Found that current chat-based AI tools struggle to meet users’ needs in jobs where workers must constantly monitor and respond to changing conditions. The findings can inform AI use-case selection and workflow design.