Patrick Neiler

Berkeley, CA · Available for Hybrid / Onsite

BUILDAI THATSHIPS.

I'm Patrick Neiler: hands-on applied AI engineer, three-time co-founder, and Co-Founder & Chief AI Officer at wrkspce. I took VoPro from zero to a signed, live enterprise pilot with Quintessential Construction Management — voice-to-operations powered by a recursive self-improvement loop where sparse expert feedback becomes a data flywheel. Try VoPro™ mini live at vopro.wrkspce.online. TypeScript · Next.js · Python · OpenAI. 15 years end-to-end: from 3E 3.0 at Thomson Reuters ($800M product line) and NBC Sports to zero-to-one at Distro AI (YC) and wrkspce.

I ship production AI products: LLM agents that do real work, not demos. I build systems that improve LLMs rather than just use them: recursive self-improvement loops where sparse expert feedback becomes a data flywheel for continuously more accurate structured output. 15 years end-to-end, three-time co-founder. Try VoPro™ mini live at vopro.wrkspce.online.

RECURSIVE SELF-IMPROVEMENT
AGENTIC SYSTEMS
DATA FLYWHEELS
VOICE-TO-OPERATIONS
DEBATE-GRAPH REASONING
CUSTOM EVAL FRAMEWORKS
SCHEMA-VALIDATED OUTPUT
TYPESCRIPT · NEXT.JS · PYTHON
HANDS-ON-KEYBOARD
LET'S SHIP
RECURSIVE SELF-IMPROVEMENT
AGENTIC SYSTEMS
DATA FLYWHEELS
VOICE-TO-OPERATIONS
DEBATE-GRAPH REASONING
CUSTOM EVAL FRAMEWORKS
SCHEMA-VALIDATED OUTPUT
TYPESCRIPT · NEXT.JS · PYTHON
HANDS-ON-KEYBOARD
LET'S SHIP

// Right now: available for new roles

Hands-on applied AI engineer building agents that get better with use.

Berkeley, CA · Available for hybrid & onsite roles

At wrkspce (founded 2025 with CEO Jeffrey Martinez), I took VoPro from zero to a signed, live enterprise MVP pilot with Quintessential Construction Management — its first production deployment on the customer's own enterprise subdomain. VoPro's recursive self-improvement mechanism turns sparse expert feedback in the field into training signal, creating a data flywheel that makes structured output more accurate with every use. Agent workflows use OpenAI function calling and JSON/Zod schema-driven validation over a comprehensive Procore schema library, converting jobsite speech into validated Procore objects (tasks, RFIs, punch items, daily logs). Try VoPro™ mini live at vopro.wrkspce.online. The pattern is reusable: the same agentic spine now powers accountability and cleaning ops, prototype to production on shared infra.

SIGNED PILOT
QCM · Live enterprise
DATA FLYWHEEL
Recursive self-improvement
15 YRS
Shipping product
3×
Co-founder
Portrait of Patrick Neiler

// Who you're hiring

Fifteen years turning ideas into shipped product.

I'm a hands-on-keyboard applied AI engineer. I own the loop from user problem to production system: prototype the agent workflow, validate structured output with Zod/JSON Schema, wire the tool calls and integrations, add the self-improvement gate, and ship the Next.js product around it. I care about amplification over automation: systems where experts inspect, fork, and refine reasoning rather than accept opaque outputs — and where every correction feeds a data flywheel that makes the next run better.

What teams hire me for: agentic workflows that survive contact with real users, schema-validated outputs that don't corrupt systems of record, and recursive self-improvement loops that turn field feedback into training signal. I move across architecture, prompt and tool design, and frontend, so the agent and the product ship together. Three-time co-founder, YC-proven shipping speed (Distro AI), speaks on applied AI at NY/LA Tech Week. A force multiplier at seed stage: I lead engineering and product simultaneously while staying hands-on in the code — an IC seat building agentic systems is exactly where I want to be.

AI alter ego

Logo of Patrick Prime

Patrick Prime is the AI manifestation of how I think and build, a working demo of agent personalization and voice-to-system interaction, not just branding.

// Track record

Shipped end-to-end, from enterprise to zero-to-one.

Applied AI infrastructure engineer and three-time co-founder with 15 years spanning enterprise legal tech (Thomson Reuters, $800M product line), sports media (NBC Sports), and zero-to-one at YC-backed Distro AI and wrkspce. Deepest in agentic systems, recursive self-improvement loops, and full-stack Next.js — agents that validate before they write to systems of record, and get better with every correction.

  • Recursive Self-Improvement
  • Data Flywheels
  • Custom Eval Frameworks
  • Debate-Graph Reasoning
  • Agentic Systems · Tool Use
  • OpenAI Function Calling
  • Zod · JSON Schema Validation
  • Voice-to-Operations
  • TypeScript · Next.js · React
  • Python · OpenAI API
  • SDK Design · Runtime-Driven UI
  • Procore Integrations

Feb 2025 to Present · United States · Applied AI infrastructure

Co-Founder & Chief AI Officer

wrkspce

01
  • Took VoPro from zero to a signed, live enterprise MVP pilot with Quintessential Construction Management (QCM), the platform's first production deployment, running on the customer's own enterprise subdomain. Try VoPro™ mini live at vopro.wrkspce.online.
  • Drove continuously improving structured-output accuracy by engineering VoPro's recursive self-improvement loop: sparse expert feedback gates convert field corrections into training signal, and a custom-built eval framework runs learning passes that turn rejected and modified drafts into rules and skill proposals.
  • Enabled automated jobsite task creation by mapping speech to Procore schemas via Zod-validated agent workflows: OpenAI function calling against a comprehensive Procore schema library turns field conversation into validated Procore objects (tasks, RFIs, punch items, daily logs).
  • Enabled multiple vertical apps (accountability.wrkspce.online, cleaning.wrkspce.online) to launch on one agentic spine by building the agent layer and UI/SDK end to end: an importable SDK with runtime schema-driven UI and swappable integration providers; zero-to-one in weeks, not quarters.
  • Established the Agentic Launchpad partnership with TheAgentic, productionizing the VoPro System & Platform Design as a deployable workspace and services offering — a strategic partnership that drives distribution beyond direct sales.

Jun 2023 to Sep 2024 · United States · two client engagements

Independent Consultant / Contractor

Carben Construction · Distro AI (YC)

02

Jul 2024 to Sep 2024 · AI product engineering · HVAC wholesale

Full Stack Engineer

Distro AI (Y Combinator)

  • Led core agent architecture for retrieval of obscure products and parts across whole manufactured HVAC systems, cutting typical part lookups for wholesale manufacturing sales teams from minutes of manual catalog search to seconds.
  • Conducted the pilot launch with realtime user feedback, triage, and resolution.
  • Developed the agent in TypeScript using LangChain, Postgres, SQL, and Next.js.

Jun 2023 to Jun 2024 · Construction procurement · multiuser workflow

Product Engineer

Carben Construction

  • Transitioned WhatsApp-based project management to a custom-built procurement system spanning multiuser workflow and AI product research for construction materials.
  • Replaced scattered chat threads with a shared, multiuser procurement workflow so stakeholders could source, review, and move material orders in one system.
  • Built AI product research into the materials workflow so the team could evaluate products inside procurement instead of hunting catalogs by hand.

Jun 2021 to Mar 2023 · Manhattan, NY · Mortgage lending SaaS

Senior Software Engineer

Simplist

03
  • Led product architecture and frontend delivery for Sonar through production launch, refactoring existing applications into a maintainable, themeable system with full configurability across 100+ Angular components.
  • Implemented WebSocket message-driven microservice architecture for reliable realtime lending workflows.
  • Developed a CSS-variables theming system as a single design-token source of truth, removing 10,000+ lines of redundant CSS.
  • Led a team of three frontend engineers to on-time delivery under tight timeline constraints.

Jun 2017 to Mar 2021 · Health-data platform

Co-Founder

Trax Health

04
  • Developed two multi-platform applications (web, iOS, Android) using Ionic and Angular for health-data monetization, delivering fully functional cross-platform MVPs.
  • Built connectivity to 170+ EHR systems and formats, enabling users to connect records from diverse providers and monetize their own health data.
  • Implemented encryption, anonymization, and consent management for sensitive health-data security and privacy, an investor-ready MVP.

Feb 2020 to Jan 2021 · Los Angeles, CA · Supply chain

Frontend Engineer

iTradeNetwork, Inc.

05
  • Rapidly designed and developed a React.js open marketplace connecting farmers and alternative buyers, with production release within 10 days during COVID-19 supply chain disruption.
  • Spearheaded a React.js component library initiative adopted across the organization, accelerating frontend development.

Jan 2016 to Feb 2020 · Greater Los Angeles · Legal tech

Lead Software Engineer

Thomson Reuters

06
  • Led development of a responsive web application (TypeScript, Angular, Redux, Azure, PowerBI) for the 3E legal billing platform through the successful 3E 3.0 launch.
  • Engineered a metadata-driven frontend framework, in partnership with the lead architect, that became the delivery pattern across the 3E platform: feature teams shipped new modules from configuration instead of bespoke builds.
  • Implemented a design system and led production of an Angular and React.js component library adopted organization-wide across the product suite.
  • Led a team of three junior engineers on legal technology products, a product line valued at $800M in enterprise sale.

Aug 2011 to Mar 2016 · Culver City, CA · Sports media

UX Developer

NBC Sports

07
  • Designed and built the NBC Sports Event Portal, a responsive Angular.js web app streamlining event planning for operations.
  • Designed 50+ interactive prototype pages and shipped 20+ features optimizing performance and UX for travel and scheduling workflows at scale.

Dec 2009 to Jan 2012 · Los Angeles, CA · Fitness SaaS

Co-Founder

Supertrainer HQ

08
  • Built and shipped Supertrainer, a business management platform for personal trainers, to production deployment.
  • Managed fundraising and investor relations, raising $75,000+ in angel investment and grants and led execution across an eight-engineer team.

2003 to 2007 · Education

BFA, Communication Design

SUNY Buffalo

09

Design foundation: why my AI products are usable, not just functional.

// What I ship

Five things teams hire me to do.

01

Agents that improve LLMs, not just use them

Recursive self-improvement loops: sparse expert feedback gates convert field corrections into training signal, creating a data flywheel that makes structured output more accurate with every use.

02

Schema-driven output that doesn't corrupt your system of record

OpenAI function calling, Zod/JSON Schema validation, and integration providers that check every LLM output before it writes to Procore, CRMs, or databases. No hallucinations sneaking into production data.

03

Voice-to-operations for real work

VoPro takes jobsite speech and produces validated Procore objects — tasks, RFIs, punch items, daily logs — via a comprehensive schema library. Live enterprise pilot with QCM; the same spine now powers accountability and cleaning ops.

04

Reasoning workbenches & custom eval frameworks

Shipped a multi-lens debate workbench (Phase 1 of the VoPro™ Reasoning Engine) where parallel reasoning agents argue over a shared debate graph with conflict detection and fork-from-claim — pressure-testing agent output before it ships. A custom eval framework wired into the product turns every rejected or modified draft into new rules and skill proposals, so expert corrections permanently upgrade the system.

05

Full-stack AI product, end-to-end

Next.js, TypeScript, React Server Components, Python, and the LLM layer. I write and ship production code every day: the agent, the API, the UI, the deployment. Zero-to-one without hand-offs.

// Labs

Apps in the labs.

A short cut of builds shipping under Patrick Prime — kitchen ops, sound maps, story stages, and the products around them.

Open labs

Flagship // Public

wrkspce.online

Construction-first voice-to-ops — speak the jobsite, land structured records in Procore

Open Site

VoPro SMB

Lightweight VoPro for small & medium teams. Speak the site walk, get validated, Procore-ready drafts without enterprise overhead.

Open App

Patrick Table

Kitchen inventory from fridge/shelf media — menus, gaps, Amazon Fresh

Open Kitchen

SoundSketch

Described songs — LLM audio analysis into black-ink maps for non-auditory listening

Open Gallery

Colby Prime

Colby Prime: openly AI artist world, twelve songs, SoundSketches, and the stage

Enter Stage

The Greatest Game Ever

Trade points, dodge Heartbreak Cop, and mint one-of-one cyberpunk story art

Play Now

Career Brain

Personal AI job agent — verified profile, fit-first discovery, truth-locked drafts

Open App

Dialectical Reasoning

Multi-lens debate workbench — provenance-tagged claims, synthesis, and forkable sessions

Open Debate

// Speaking & community

On stage and in the room.

I speak on what actually ships in applied AI: production agents, eval, and how teams move from prototype to revenue without rebuilding. Recent talks and hosted events below. Booking for panels, firesides, and workshops in SF/Bay Area and beyond.

YC Reject Pitch Night · SanscriticPitcher · VoPro by wrkspce.online

Recursive Self-Learning Pitch Event

Jul 2026 · Workato HQ

Live VoPro demo at YC Reject Pitch Night: speak into the dock, capture intent, and rebuild the deck from what you say. A room of builders proving rejection isn't a verdict, hosted by Sanscritic at Workato HQ.

New York Tech Week 2026Panel speaker

The Impact Room

Jun 2026 · New York

Panel on where agentic systems are genuinely useful vs. hype and how product teams should prepare for a world where describing the outcome is enough and the system does the work.

Tech Week LA 2025Host

AI on the Rocks

2025 · Los Angeles

Hosted LA Tech Week event bringing builders and operators together to trade what's actually working in applied AI, less keynote, more real conversation.

Book me for your event.

Panels, firesides, workshops, or hosting: if it's about shipping AI that earns its place in real work, I'm in. Local to Berkeley/SF, available for remote events.

Book Patrick
Patrick Neiler: Staff Applied AI Engineer, Berkeley, CA

// Why hire me now

One vertical done right, then the platform scales.

I proved the model with construction, the highest-friction, lowest-trust test for AI. The same agent spine now spins up new verticals in weeks. If you're a team that needs a Staff Applied AI Engineer who ships production agents and the product around them, this is the hire that shortens your roadmap.

  1. Now

    Live pilot → measurable outcomes

    VoPro in the field with QCM, instrumented for time saved, validation rate, and intervention rate. Building the eval dataset real products need.

  2. Next

    Harden the agent platform & SDK

    Ship the eval harnesses, schema library, and integration providers as importable infrastructure, so your team gets production agents without starting from zero.

  3. Ahead

    Extend to new domains with you

    Apply the spine to your domain, whether it's field ops, back office, or customer workflows and create a defensible product where competitors are still demoing.