Chad Harmer · Product leadership

Product leadershipwhere workflows, data and AI meet.

Fifteen years across product management, process improvement and enterprise systems at Capital One, PNC and TD Bank progressed into hands-on AI product development. Since 2025, I’ve taken production AI products from discovery through architecture, evaluation, launch and iteration.

See selected workExperience
four production AI products across recruiting, education and consumer apps
Web + iOS
standards records across 185 frameworks in 17 states, grounding retrieval · VeraTeach
110K+
rubric evaluation after a measured model change, ~2.1× cheaper with scores held · VeraTeach
74s → 19s
consumer subscription products
Paying customers
Background

Product leadership built from understanding how the work actually happens.

My career has progressively moved closer to the systems behind the customer experience — from customer-facing work and process improvement, to owning enterprise data-management processes, to product management and software delivery.

At Capital One, I moved from owning and improving enterprise data-management processes into product management, leading a cross-functional Agile delivery team of engineers, designers and data analysts to build software that simplified complex validation and control workflows.

Through Yellow Labs Products, that progression moved into hands-on AI product development, taking production products through discovery, architecture, implementation, evaluation, launch and iteration.

The progression has been consistent: understand how work actually happens, improve the system around it, and increasingly build the technology that makes it better.

Capabilities

Capabilities in practice.

Product leadership, AI systems, evaluation and data — demonstrated across enterprise delivery and shipped products.

A

AI product systems

Grounding · RAG · orchestration · structured outputs · model / code boundaries · human-in-the-loop

  • VeraTeachRetrieval-grounded generation over 110,000+ standards records, abstaining when the supporting evidence isn’t strong enough.
  • RygerEvidence-grounded candidate matching and resume generation, constrained to experience the candidate can actually support.
  • ReadMyStripA clear model / code boundary: a vision model reads the test strip, and deterministic code validates every reading and makes the call.
  • VeraTeachHuman-in-the-loop by design: AI drafts lessons, materials and grading suggestions, and teachers approve every score before it reaches Google Classroom.
B

Evaluation & AI quality

Evals · model and prompt comparison · regression · quality / cost / latency · reliability

  • VeraTeachModel choices evaluated blind across quality, latency and cost; one measured change cut evaluation from 74s to 19s at ~2.1× lower cost with quality essentially unchanged.
  • VeraTeachLesson quality scored against an explicit rubric, and regression-checked against real models when prompts or models change.
  • ReadMyStripReliability on real-world photos: image quality is checked before any model call, and a weak scan prompts a retake rather than a partial answer.
C

Data & decision intelligence

Structured data · normalization · evidence · matching · classification · deterministic logic

  • Grocery SidekickTurned messy source data into a normalized, deterministic product-intelligence layer across 400,000+ foods and 430,000+ barcodes, powering barcode lookup, label scanning and product alternatives.
  • VeraTeachBuilt a standards corpus of 185 frameworks across 17 states, sourced from each state and used as a shared foundation for lesson planning, imported materials and assessments.
  • RygerStructured candidate intelligence, built once per candidate, supports both recruiter matching and resume generation — with Met / Review / Missed assessments a recruiter can check.
D

Product leadership & delivery

Discovery · strategy · workflow design · cross-functional leadership · measurement · shipping

  • Capital OneLed a cross-functional Agile team across engineering, design and data to turn a complex manual validation workflow into an enterprise platform, reducing ~57 user interactions to 2.
  • Capital OneSet the vision and roadmap for a new enterprise product with senior leadership, and started cross-product PM collaboration so teams shared features instead of rebuilding them.
  • RygerRecruiter research revealed that the bigger problem wasn’t finding one match, but losing valuable candidate relationships — shifting Ryger from a matching tool toward a recruiter-owned talent network.
  • Yellow LabsShipped four production AI products across web and iOS, including subscription software with paying customers, from discovery through launch and iteration.
Selected Work

Different domains. The same product discipline.

My work spans enterprise data management, recruiting, K–12 education, food intelligence and water testing. Across each, the work starts by getting deep into the users, workflows, data and constraints — then turning that understanding into the right product and system.

VeraTeach

Live · Web · K–12

An AI teaching workspace that carries a lesson from state standards through Google Classroom to returned grades.

  • Multi-stage AI orchestration
  • Shared 110K+ standards layer
  • Evaluation and model routing
  • Human-in-the-loop grading
Key decision
AI drafts grading suggestions; only teacher-approved scores go back to Classroom.
Signal
Rubric evaluation went from 74s to 19s at ~2.1× lower cost, with near-identical scores.
Read case studyverateach.com ↗

Grocery Sidekick

Live · iOS

Scan a grocery product to see which of the allergens and ingredients you watch for it contains, and which ingredient is responsible.

  • Large-scale data normalization
  • Ingredient resolution
  • Deterministic classification
  • Benchmarks and audits
Key decision
Turn messy source data into a reliable product layer: inconsistent label records normalized into stable product identities, ingredients resolved against a curated dictionary, and classifications computed deterministically at build time.
Signal
400,000+ foods and 430,000+ barcodes bundled on the phone, so lookups work with no signal.
Read case studyApp Store ↗

Ryger

Live · Web

Recruiting intelligence with evidence as the boundary: it structures candidate experience, matches it against role requirements, and generates targeted resumes limited to what the candidate can actually support.

  • Requirement-level matching
  • Evidence-grounded resume generation
  • Shared candidate-intelligence layer
  • Discovery that reshaped the thesis
Key decision
Evidence is the boundary. Each requirement is Met, Review or Missed with its supporting evidence, and a generated resume line whose numbers, dates or employers aren’t in the source is removed rather than trusted.
Read case studyryger.app ↗

ReadMyStrip

Live · iOS · Paying customers

Photograph a pool or spa test strip next to your own bottle’s chart and get every reading in seconds.

  • Vision-model pipeline for real-world photos
  • Deterministic validation of model output
  • Subscription monetization
  • Iteration from production feedback
Key decision
The model only perceives; code decides. A reading that isn’t printed on your chart becomes “unknown,” never a guess.
Signal
Paying customers — and production feedback led to scan history and notes.
Read case studyApp Store ↗
All work and how it maps to capabilities
Experience

Fifteen years of systems work. The tools changed.

My career has been spent on the workflows, controls and data underneath large organizations — from customer-facing work and team leadership, through process documentation and process management, into enterprise data management and product. The problems stayed consistent: where work gets stuck, what can’t be measured, and what people can’t trust.

Education & certifications

  • B.A. Psychology, Rutgers University
  • Certified Scrum Product Owner (CSPO)
  • AI for Product — Product School
  • Product Leader Certificate — Product School
  • Lean Six Sigma Green Belt

  1. Yellow Labs Products

    January 2025 – Present

    Founder / AI Product Builder

    An independent product studio building and operating production AI web and mobile products in recruiting, education and consumer apps. I own the lifecycle from discovery and market research through strategy, experience design, architecture, implementation, launch, measurement and iteration.

    Selected work

  2. Capital One, Risk Management

    October 2018 – September 2023

    Manager, Product Management

    October 2020 – September 2023

    • Owned strategy and delivery for VaaS, a platform supporting data-management validations and control testing.
    • Led an Agile delivery team of engineers, designers and data analysts as product manager, partnering with engineering, validators, risk stakeholders and enterprise teams to turn control and process requirements into technology.
    • Led the move from a Collibra-based front end to a standalone AWS-hosted application; automation cut a manual validation workflow from ~57 user interactions to 2.
    • Increased high-priority usage validations completed per quarter by 15%.
    • Built the roadmap, strategy and vision for a new product with senior leadership, and started cross-team PM collaboration to share features and reduce arbitrary differences between products.
    • Designed rollout timelines and led training for frontline users.

    Manager, Process Management

    October 2018 – October 2020

    • Redesigned enterprise data governance from a consumer-driven Data Usage Agreement model to a producer-driven Data Provisioning Agreement model, removing duplicate requirements and making accountability scale.
    • Built reporting and KPIs where none existed; used them to trace low completion of a governance activity to a ~4-month approval backlog, then eliminated the backlog.
    • Led design and delivery of the Effective Challenge Guidelines, helping first-line users understand data-management requirements.
    • Subject-matter expert across enterprise projects; mentored newer process managers.

  3. PNC Bank

    2017 – 2018

    Process Improvement Consultant

    Coordinated cross-functional problem solving using PDCA, ran root-cause analysis on service metrics, and coached managers on performance huddles built around metrics and continuous improvement.

  4. TD Bank

    2008 – 2017

    Business Process Manager

    Five roles over nearly a decade, from frontline customer service through team leadership, contact-center process documentation and retail-banking process management. Led a Lean Six Sigma Green Belt project on coin reconciliation that contributed to an estimated $2.5M in savings; built a decision-support dashboard modeling TD and competitor operating hours by market; joined TD’s first social-media team and led a group of 10 agents.

How I Work

Principles, with the decisions behind them.

How I make product calls — each one anchored to a decision I actually made, not a slide.

01

Find the bottleneck before building the fix.

Low completion looked like an adoption problem. The data showed a four-month approval backlog.

Capital One · Ryger

02

A confident wrong answer costs more than no answer.

“Couldn’t tell,” “Review” and “unknown” are product features. When an evaluation benchmark turned out to reward confident invention over asking a clarifying question, I changed the benchmark.

Grocery Sidekick · Ryger · VeraTeach · ReadMyStrip

03

AI drafts; accountable people decide.

Where someone answers for the outcome, the model suggests and the person decides.

VeraTeach grading · control testing

04

Remove steps before adding features.

Taking a workflow from 57 interactions to 2 did more than any new capability would have.

Capital One

05

Measure, then decide — including deciding not to ship.

Model routing set by blind comparison. Fuzzy matching measured and turned off. KPIs built where none existed.

VeraTeach · Grocery Sidekick · Capital One

Contact

Hiring for AI or platform product leadership?

I’m looking for Senior and Principal product roles in AI, platform and technical product management — teams where product judgment and a working understanding of AI systems both matter. Email is the fastest way to reach me.

chadharmer@mac.comLinkedIn