US-based · Open to opportunities

Applied AI and enterprise APIs.

I’m Kevin Houston, an applied AI architect and solutions engineer with enterprise experience at Meta, Yahoo, and Criteo. My current consulting spans production multi-agent systems, customer-data architecture, evaluation, and training.

bounded-ai / production client
sources→ approved facts + selected live data
model roles→ zero, one, or two calls per request
application→ validation, responses, destinations
operations→ evaluation, diagnosis, escalation
> models propose approved facts and actions
> application code controls the response
01
Production AI delivery
Client discovery, bounded model roles, evaluation, and live training.
02
10+ years of technical delivery
Customer-facing discovery, enterprise integrations, and enablement.
03
Supporting $10M+ in portfolio revenue
Fortune 100 server-to-server integration architecture at Yahoo.
Prior platforms
Meta Yahoo Criteo
01Selected work

Production delivery and the systems behind it.

Client implementations, API architecture, and self-directed tools. Each case explains my role, the implementation decisions, and what was verified.
Production deploymentClient delivery · 2026

Bounded AI & Customer-Data Architecture

A bespoke, customer-facing multi-agent system for a services-business client, from discovery through production support.

Mapped website, scheduler, CRM, and contact workflows. Designed application-enforced controls, passed 46/46 authored development scenarios in the final evaluation run, recovered production service, and delivered illustrated guides and live one-on-one training.

TypeScriptEvaluationCustomer dataClient enablement
  1. Approved business sourcesMultilingual facts and selected live scheduler prices/provider listings.
  2. Bounded model decisionsTwo conditional roles propose approved fact IDs and actions.
  3. Application validationSchema and allowed-value checks; controlled wording and booking/contact destinations.
  4. Customer response or escalationApproved information, a permitted next step, or a route to a person.
Active Featured Mar 2025, present

AI Operator System

A file-and-review workflow for keeping AI-assisted development scoped, reviewable, and easier to resume across sessions.

Reusable documentation and templates separate project rules, specs, review checks, and handoff state. Claude Code and Codex can read the same repo context; the maintainer checks the work and decides what ships.

Spec-before-build Human approval gates Claude Code Codex
Released application

Source

Local-first raw transcript capture

A local macOS app that captures mic, system audio, or both at once, and transcribes the audio on-device with faster-whisper. Existing meeting tools optimize for summaries. I wanted the raw transcript source file, saved locally as Markdown with YAML frontmatter and per-source WAV sidecars, so I can hand it into whatever tool I reach for next.

Python faster-whisper Local-first

Installation and setup instructions are available in the project README.

Reference implementation

Support Triage Starter

Rule-based support triage

A FastAPI reference implementation that classifies three issue types, asks at most one clarifying question, and returns structured routing decisions. Uses deterministic rules and mock tools, with explicit escalation paths.

Structured output FastAPI Escalation logic
Additional work Four more projects

Opportunity Pipeline

Human-in-the-loop workflow · Public documentation

A personal AI-assisted workflow for opportunity discovery, filtering, scoring, and approved artifact generation. The public documentation explains suppression, review gates, and operating constraints; source code and personal data remain private.

WaveAIO

Archived product · 2021–2023
On-chain minting and automation

Co-founded and led an NFT minting platform end to end. Owned product direction, GTM strategy, contributor coordination, and customer-facing execution from concept through launch.

Circle CCTP

Testnet integration · Nov–Dec 2025
ethers.js · USDC

Implemented the full burn/mint cycle across Base and Ethereum Sepolia via Circle's Cross-Chain Transfer Protocol. Hands-on work with cryptographic attestation and USDC settlement mechanics.

Uniswap V3

Strategy document · Mar 2025
Developer support and API strategy

Designed a tiered SLA escalation model and phased self-service transition roadmap for the trading API. Covered swap routing, liquidity-pool queries, and multi-channel developer triage.

--Writing

Notes on building in practice.

Published April 2026

Building Source, a Local-First Transcriber for Raw Transcript Ownership

I did not build a new transcription model. I built a local-first wrapper around faster-whisper because the existing meeting tools optimize for summaries, and I wanted the raw transcript source file. The writeup covers macOS packaging, mic and system audio capture, model tradeoffs, and what I learned about input quality versus model size.

Read the Source build notes
02How I work

Four steps, the same loop from discovery to ship.

My focus is the work around the model: defining scope, checking outputs, handling exceptions, and making ownership clear. Enterprise API experience informs how I approach each AI workflow.

Clients don't pay for APIs. They pay for the outcome the integration makes possible, and for someone who'll see it cleanly through the launch.

Step 01
01

Discover

Map the system, the stakeholders, and the constraints before proposing anything. Most SE failures are scoping failures in disguise.

Step 02
02

Spec

Write down scope, acceptance criteria, and the exceptions worth handling. The integration, not the API, is the product.

Step 03
03

Review

Human approval at the gates that matter. Suppression rules, exception paths, and fallbacks, not just the happy path.

Step 04
04

Ship

Optimize for the handoff as much as the launch. Documentation, ownership boundaries, and repeatable patterns are part of the deliverable.

03Experience

Enterprise experience, applied to current delivery.

Discovery, integration design, launch support, and technical enablement across enterprise platforms and independent client work.
Apr 2025 to present
United States

Applied AI Solutions Architect | Multi-Agent Systems

Lead client discovery, customer-data architecture, and applied AI delivery. Designed and deployed a production TypeScript system with two bounded model roles, application-controlled responses, evaluation and operational tooling, and human escalation. Delivered bespoke guides and live one-on-one training.

Production AI
Architecture through support
Customer-data discovery
Website, scheduler, CRM, consent
Jun 2023 to Feb 2025
Remote

Technical Solutions Consultant (contract)

Lead technical consultant for enterprise API integrations on Meta's advertising platform. Drove adoption through developer enablement, integration playbooks, and cross-functional product feedback loops across Conversions API, Signals, and Measurement surfaces.

Pixel / Conversions API
Measurement troubleshooting workshops
~90 accounts/month
Enterprise advertiser API and measurement support
Nov 2020 to Feb 2023
Los Angeles, CA

Solutions Engineering Manager / Solutions Sales Engineer

Served as the DSP API subject-matter expert for entertainment advertisers. Led one engineer in building the DSP onboarding implementation, and guided enterprise discovery, server-to-server integration design, launch planning, and reusable API playbooks.

Supporting $10M+ in portfolio revenue
Fortune 100 S2S integration architecture
30-35% shorter time to launch
API playbooks and hands-on implementation guidance
Dec 2018 to Nov 2020
New York, NY

Solutions Engineer

Owned technical discovery, pre-sales solution design, and integration strategy across e-commerce partners. Full-stack integration consultant across API and catalog-driven personalization pipelines.

$3M week-over-week revenue lift
Contributed through Shipt integration diagnosis
45% faster implementation
Across six retail media API launches
Education

Engineering foundations, applied to delivery.

Master of Engineering

Information Engineering Management & Technology Commercialization

University of Alabama at Birmingham

B.S., Electrical & Computer Engineering

University of Alabama at Birmingham

04Technical surface

Where I'm hands-on.

Enterprise integration and customer-data experience, with current hands-on work in bounded AI systems, evaluation, and operating controls.
01
Applied AI & Evaluation

OpenAI API, bounded model roles, structured outputs, schema validation, authored-scenario LLM evaluation, and software regression checks.

02
Customer Data & Delivery

Website/scheduler/CRM discovery, consent and data minimization, enquiry attribution, incident diagnosis, operating runbooks, and live client training.

03
Enterprise Ad Tech

Yahoo DSP API, Meta Pixel/Conversions API, Criteo Retail Media, server-to-server events, CDPs, tag managers, identity, mobile measurement, and offline conversions.

04
Languages & Tools

JavaScript/TypeScript, Node.js evaluation and operational tooling, SQL/Vertica, REST APIs, OAuth, Postman, Git, and Python/FastAPI prototypes.

05Contact

Let's work together.

Open to Applied AI Solutions Architect, Solutions Engineer, Solutions Architect, and Forward Deployed Engineer roles. US-based, with Los Angeles as my primary location target. Open to remote or hybrid work and relocation for the right role.