James R. Weaver / Applied & Generative AI / Architecture to Working Solutions
Three decades leading software organizations. Today a Prototyping & AI Cloud Engineering leader at AWS, converting enterprise generative AI from architecture into working solutions. Strategy you can operate, not just present.
Three decades leading software organizations, and still shipping.
Always-on agents don't pencil out on metered inference. That's the whole argument for owning the silicon. A dozen local models across Mac and DGX-class machines behind one endpoint, smart-routed, with memory-aware scheduling that accounts for what parameter counts don't tell you.
In production Exhibit B / Agentic SDLCAsk a model for a whole application and it truncates. Ask it to fix one file and it breaks two more. Forge is what's left after designing around both: one file per LLM call, surgical patches instead of regeneration, and telemetry on what the machines got wrong.
Active development Exhibit C / MemoryAn agent that forgets every session can't be delegated to, and the unlock isn't a bigger model, it's memory the whole team shares. Persistent semantic memory on PostgreSQL and pgvector: decisions and learnings that outlive the session, read by every agent in the fleet.
In productionFour departments: Musings on all things AI · The White Rabbit, experiments followed wherever they go · By the Numbers, concrete data over vibes · Counterpoint, my take on the discourse.
A technology executive who still ships, because strategy you can't operate is just opinion.
Across three decades I've led product and engineering organizations across financial services, casino gaming, healthcare SaaS, medication-management analytics, public-sector modernization, and now a Prototyping & AI Cloud Engineering leader at AWS. Today I take enterprise generative AI from architecture to production launch. After hours I run a production inference fleet and a team of autonomous agents, because the questions that matter can't be answered from slideware.
Judgment, allocated three ways.
Correspondence
If you're wrestling with enterprise AI transformation, chasing your own white rabbit, or want to compare notes on what the data actually says. My inbox is open. Board inquiries: same inbox, say so in the subject.