ABOUT
Peter C. Bennett
Forward-deployed AI engineer. I turn ambiguous enterprise problems into AI products teams can operate.
- Looking to hire
- Open to Principal AI Solutions Architect, Forward-Deployed AI Engineer, and senior enterprise AI delivery roles. The résumé is the fast read, and the five case studies tell the story behind each line.
- Scoping a project
- Discovery through production, for AI systems that have to live inside a real workflow. An email starts it, and the method pages show how an engagement runs.
- Technical peer
- Two releases are public: CaseFlow, the 21-node LangGraph workflow with its eval suite and human gates, and Local Workspace MCP, a read-oriented MCP server with a published threat model. Both publish their source.
- Pressed for time
- Three stops: the résumé for the career, the public releases for the engineering, the method pages for how an engagement runs.
THE ARC
Three decades inside the systems that run businesses.
My career runs through the unglamorous core of enterprise software: the systems that pay people, integrate acquisitions, move product, and file the reports. That background is why the current work is about AI that survives production: most of the difficulty sits in everything the model has to live inside.
- Early enterprise systems
- Advanced Systems Consulting: systems for GE Corporate, GE Information Services' billion-dollar B2B marketplace, and Equitable Resources.
- Executive systems
- Conversion Services / Bristol-Myers Squibb: HR compensation and executive stock-planning systems serving 17,000+ executives.
- Bioinformatics
- Accugenix via Tech USA: a bioinformatics portal for 3,000+ pharmaceutical clients that enabled a $25M+ revenue line.
- Hearst Magazines, 2007–2018
- Senior web application and API developer, then technical integration lead on a $10M+ social-platform acquisition. APIs behind subscription, content, and identity systems tied to $50M+ recurring revenue and 1M+ monthly visitors.
- TTAC Publishing
- The executive team's technical owner across three ecommerce properties. Recurring AWS cost cut by more than 90%, 99.97% uptime, roughly triple the release cadence.
- Forward-deployed AI engineering
- Two years of embedded AI delivery: the LeadsON operating platform, plus two public releases, CaseFlow (a reference implementation of human-accountable agentic workflow) and Local Workspace MCP (a read-oriented MCP server shipped through adversarial review).
THE LANE
What the work is now.
Forward-deployed AI engineering: embedding with the people who own a problem, writing down how the work runs, and shipping bounded AI systems that prove themselves inside real workflows before they are called done. It is the pattern behind CaseFlow, LeadsON, and the client work before them.
Areas of practice: technical discovery, agentic workflow design, typed and validated AI boundaries, evaluation suites, systems integration, and the production signals that let a team trust what shipped.
KEEP READING