David Leathers

Applied AI · Technical Product · Revenue Systems

I design AI systems that do real work.

I'm David Leathers, an Applied AI and technical product leader who turns difficult revenue and operational problems into production software, governed agentic workflows, and marketplace infrastructure. My background spans AI-native product development, performance marketing, telephony, real-time marketplaces, and executive operations.

View selected work → Download résumé

Tampa Bay, FL · Relocating to Dallas–Fort Worth, TX

  • 200 Registered marketplace organizations 50 registered buyer organizations and 150 registered publisher organizations.
  • 10K–13K RTB requests on an active day
  • $15M Marketing budget overseen
  • 88+ Documented AI-operations engagements in 10 weeks

Selected work

Systems, not side projects.

Production marketplace

Dependable Call Exchange

A multi-tenant performance marketplace connecting buyers and publishers through real-time bidding, telephony, routing, attribution, billing, payouts, partner workflows, and production AI.

Role
Founder & Product Lead
Evidence
Private production system · sanitized case study
  • 200 Registered marketplace organizations 50 registered buyer organizations and 150 registered publisher organizations.
  • 30 Campaign verticals represented
  • 10K–13K RTB requests on an active day
  • 70+ Configured buyer and publisher RTB integrations

Read the case study →
03 AI operations

Multi-day operations reduced to minutes of human oversight

A governed AI operations layer handles finance, reconciliation, analytics, investigations, and draft business workflows while preserving human authority over external communications and money movement.

Role Operating model & guardrails

See how the operating model works →

  • 88+ Documented AI-operations engagements in 10 weeks
  • $48.4K Total invoiced $48,423.37 across 19 invoices.
Private production system · sanitized case study
The AI-native product-development loop: eight steps that close on themselves A closed eight-step cycle. Read clockwise from the top: 1, understand the business problem; 2, research until i can challenge the solution; 3, define the prd, requirements and acceptance criteria; 4, direct coding agents through implementation; 5, review and test adversarially; 6, deploy deliberately; 7, observe production failure modes; 8, encode what was learned into permanent rules and controls. The eighth step returns to the first along a dashed arc labelled "encoded and reused", which is what makes this a loop rather than a list — the loop closes when a production failure becomes a permanent control. The connecting spine is brass because brass marks the primary control path. Steps four and five, the two steps where a coding agent is on the keyboard, carry verdigris node rings because verdigris marks agent pathways. Step eight is a filled brass terminal because closing the loop is itself the control point. encoded and reused Understand the business problem 1 Research until I can challenge the solution 2 Define the PRD, requirements and acceptance criteria 3 Direct coding agents through implementation 4 Review and test adversarially 5 Deploy deliberately 6 Observe production failure modes 7 Encode what was learned into permanent rules and controls 8
Eight steps. The loop closes when a production failure becomes a permanent control.

How I work

The agent writes code. I own the outcome.

  1. Understand the business problem.
  2. Research the system until I can challenge the solution.
  3. Define the PRD, requirements, constraints, and acceptance criteria.
  4. Direct coding agents through implementation.
  5. Review and test adversarially.
  6. Deploy deliberately.
  7. Observe production failure modes.
  8. Encode what was learned into permanent rules and controls.

I don't position myself as the engineer who manually authors every line of production code. My role is to define what the system must do, understand the underlying technical concepts well enough to challenge implementation choices, direct AI-native delivery, validate the result, and own what happens in production.

How I build with AI →

Applied AI

AI that does real work — not demos.

Three production systems where an agent's output changes something real — and where the interesting engineering is in what happens between the model's suggestion and the system's state.

  • Scoped permissions
  • Two-turn confirmation
  • Stale-state protection
  • Idempotent execution
  • Audit logging
  • Verified outcomes
  • Explicit human escalation
  • ~35 Partner Chat tools

    Partner Chat

    Natural-language partner support and production configuration actions, behind scoped permissions and explicit confirmation.

  • ~120 Typed operator MCP tools

    Operator MCP

    Governed internal agent access to administrative workflows across the platform, every tool typed and audited.

  • 88+ Documented AI-operations engagements in 10 weeks

    DCE-Ops

    Finance, reconciliation, analysis, investigations, and operational support — with money movement reserved to a human.

Leadership & scale

Technology is useful because businesses have outcomes.

  • $2M+ Monthly paid media directly managed Sustained for 6+ months.
  • 15K+ Daily billable inbound calls Generated by programs under David's leadership, including affiliates in his network.
  • 10K+ Policy sales supported During a two-month Open Enrollment period.
  • ~200 Partner and vendor organizations managed

Before building AI-native software systems, I operated at executive scale in performance marketing. That experience is why I care about economics, attribution, routing, conversion, operational reliability, and whether a system produces a business result — not whether the technology sounds impressive.

Leadership & scale →

Writing

Technical notes

  1. 01

    The Most Valuable AI Output Was a Refusal to Compute

    How a symmetric billing bug made a losing campaign look profitable — and why a production AI system should sometimes refuse to produce an answer.
  2. 02

    What Building Production Software With Coding Agents Actually Requires

    The parts of AI-native delivery that are not the code: requirements, acceptance criteria, adversarial review, and owning the production outcome.
  3. 03

    Agent Confirmation Is a Product Problem, Not a Button

    Model output is not authorization. What it actually takes for an agent to change real system state safely.

Read technical notes →

Building something where AI has to be useful, accountable, and economically meaningful?

I'm interested in senior Applied AI, technical product, AI transformation, revenue-systems, and related leadership opportunities — especially roles where software has to survive contact with real operations.

Discuss an opportunity LinkedIn