30 years building and running the systems that carry a company's money and the systems that carry its risk. Today that means production AI: model routing that scores cost against quality in real time, spend governance baked into the request path, and multi-agent systems running live, not in a demo. Scale, reliability, accountability. That triad is the operating discipline behind everything I build.
I do not advise on AI from a deck. I build it, every day. Most AI spend today is ungoverned, and one bad month from becoming an emergency. A developer who could never expense a $25 lunch is spending $10,000 in tokens with nobody checking. That gap is what I build the systems to close.
“Jerry is one of those rare IT executives who combines deep technical expertise with an incredibly positive, energetic, and people-focused approach. He elevates the culture around him, builds trust quickly, and consistently follows through…” more
Working with AI tools every day shapes a distinct point of view. These are the principles that guide how I evaluate, deploy, and lead AI initiatives.
AI-powered change and competitive pressure will demand rapid reductions in technical staffing across the entire lifecycle — planning, development, QA, security, documentation, deployment. Only leaders who are hands-on enough to truly know what is possible today will have the clarity to make the right calls. If you are not immersed in AI tool changes daily, you cannot keep up.
Models change every day. Tools change every day. Abstractions shift. Metrics shift. The old model of layered technology management with distance from the line is dying. What is coming demands intimate knowledge of what shifted today, how fast it is moving, and where the acceleration is heading.
You cannot pick one AI vendor and one model for all things. I run my own development work through a personal six-seat review board. One model per vendor, each seat assigned a different area of scrutiny, before anything reaches QA. Genuinely different vendors produce genuinely different judgment. Any organization that has not built that kind of cross-model discipline into its own workflow is already behind.
The new paradigm is agent teams. Each team has a sphere of responsibility. Each agent is specialized via background and persona definitions. These specialties need a model precisely matched to their needs — a public model with per-token cost, a local model for specialized tasks, or a mix tuned to each outcome.
Not every task needs a frontier model. A summarization job does not need the same firepower as a complex reasoning chain. I built a least-cost routing engine that scores 64 active models across 16 providers on cost, quality, and reliability — then auto-selects the cheapest model that meets the task requirements. The savings compound fast when every request is matched to the right tier.
Tokens are cash, and I treat them that way. Every AI request in my systems is logged, attributed to a client, and measured against a budget before the call is even made. Cost governance is not a report you check after the bill arrives. It is baked into the routing layer from day one.
When every team and every agent can call an LLM, you need a single chokepoint that enforces the rules. My proxy centralizes authentication, rate limits, audit logging, and usage policies — no individual service touches a provider API directly. This is not optional governance bolted on later; it is the architecture. Every prompt in, every response out, fully attributed and auditable.
“I got to know Jerry as a capable leader, able to motivate involved staff and someone who can get the job done fast and effectively. I'd highly recommend Jerry for his spirited leadership…” more
Not a list of things I have heard of. These are tools and models I evaluate, compare, and use in production — daily.
Not demos or proofs of concept — these are live systems processing real data, managing real money, and routing real API traffic every day.
Multi-provider LLM routing with cost optimization
Centralized API that catalogs 64 active models across 16 providers (Anthropic, OpenAI, Google, xAI, Meta, Microsoft, DeepSeek, Mistral, Perplexity, Cohere, Amazon, Alibaba, Zhipu, Moonshot, MiniMax, ByteDance), routes requests via least-cost scoring, and tracks per-model spend. Built-in cost-management layer enforces budget caps, surfaces cost-per-token analytics, and auto-selects the cheapest capable model for each task — cutting LLM spend without sacrificing output quality.
Policy-enforced async gateway for centralized AI access
SQS-based async service on AWS that receives AI prompts from any client, selects the optimal model via the router, and returns results. Enforces per-provider rate limits, request-level authentication, and usage policies centrally — no individual service touches a provider API directly. Includes retry logic, dead-letter queues for failed requests, FIFO ordering guarantees, and full audit logging for cost and security compliance.
Multi-strategy portfolio automation with LLM-powered analysis
End-to-end trading platform managing a live brokerage account through 10+ strategy plugins. A 4-agent LLM swarm (technical, fundamental, sentiment, macro) analyzes 80 tickers nightly, synthesizes buy/sell signals, and feeds a conviction-weighted portfolio engine. All state in PostgreSQL + TimescaleDB with real-time dashboard.
The SIP switch built by AI, for AI
Replaces traditional active/standby SBC pairs with a swarm mesh of identical, expendable nodes. Each node runs 10 LLM agents handling configuration, routing, security, and fraud detection in real time. A bid/ask pricing engine — modeled on stock exchange mechanics — negotiates rates per-session across AI model traffic, voice/video, and SMS. Hash-chained CDRs with Merkle tree verification provide financial-grade audit trails. Zero proprietary protocols.
“He has a keen understanding of all of the technology involved in any project and is usually the person that provides directional design. From budgets to contracts or development and everything in between…” more
Recommendations from colleagues, direct reports, and executives across multiple organizations.
Jerry was a great partner to the finance team. He took the time to dig deep and understand the existing data and structures to support the analytics team in how best to approach questions and report development. Jerry also made data easily available to finance to enable better understanding of the business. He was always willing to roll up his sleeves and do detailed research and analysis. more
System and method for normalizing unstructured text data across disparate formats and sources for downstream processing and analysis.
System and method for intelligent data targeting and routing based on real-time pattern recognition and classification rules.
I'm open to VP, CIO, CTO, and senior director-level opportunities in AI transformation, cloud modernization, and enterprise technology leadership. Available for both full-time executive roles and advisory engagements where daily, hands-on AI expertise drives immediate operational impact.
Denton, TX