peopleanalyst

CV

On-page reading version. Use ↓ Save as PDF for a print-styled copy, or print from your browser.

Mike West

Product & Technical Leader · AI-Native Analytics Platforms · Applied Decision Science Pittsburgh, PA · mike@peopleanalyst.com · 469.406.4699 linkedin.com/in/michaelcwest · github.com/people-analyst · peopleanalyst.com

Positioning

Mike West wrote the first mainstream book on people analytics (People Analytics For Dummies, Wiley 2019) and built the function from zero at Google (through 7K → 21K growth), Merck (65K employees, $4B HR investment), and PetSmart. He is the only person to have worked on three of the field's leading platforms — Visier, One Model, ZeroedIn.

Since 2022 he has been building the case for an inversion. Most of the field argues that AI improves analytics. But an AI that judges people is a rater — and a century of psychometrics already knows how to make a noisy rater reliable. The under-told direction is that behavioral science directly improves AI, and the portfolio is the argument.

The argument ships. 45+ live analytical services, exposed over HTTP for engineers and MCP for AI agents — all 90+ tools agent-callable, enforced by an automated check rather than promised on a roadmap. A corpus engine that turns books into queryable data with provenance on every claim. Three provisional patents (2026), each claiming a system already running in production. And posted prices on a live till — because analytics should be a decision you buy, not an engagement you negotiate.

Seeking founder-track placement (South Park Commons, Entrepreneur First, On Deck, YC) or senior product leadership (Principal / Staff PM, VP Product, Head of Product / AI) at AI-native and applied-analytics companies. Concurrently applying to PhD programs at Carnegie Mellon (HCII) and the University of Pittsburgh.

Distinctive Claims

  • First person to have worked on three leading workforce-analytics data platforms — Visier, One Model, ZeroedIn.
  • Author, People Analytics for Dummies (Wiley, March 2019) — first mainstream book on people analytics. 22K LinkedIn followers, 97+ published articles, group moderator of the People Analytics Community.
  • Solo founder of an AI-native TypeScript ecosystem spanning enterprise analytics (People Analytics Toolbox — 45+ live microservices over HTTP + MCP), corpus + factory infrastructure (CanonicAI engine), performance intelligence (Performix), scholarly research surfaces (Principia, Namesake research apparatus, Penwright Research Program), and consumer products (Vela, Namesake, Bicycle Guide, Fourth & 2). 2.6M+ lines across 20+ active repositories; 16,000+ commits in the last twelve months — measured from the GitHub org, not estimated, and growing weekly.
  • Shipped the revenue infrastructure, not just the analytics — a live Stripe till carrying 40+ posted-price SKUs (datasets, feeds, assessment kits, metered API tiers) across the portfolio, on a posted-pricing model rather than quote-based consulting.
  • Multi-year product leader (Strategist / Manager / Principal Engineer) at people-data startups (OneModel, OpenComp, AnyComp.AI, Performix) running concurrently with consulting practice.
  • Originated methodologies adopted in the field: Rapid Collaborative Insight (RCI), Net Activated Value (NAV), Three A's Framework, Lean People Analytics, Full-Stack People Analytics Systems, Quantitative Model of HR.
  • Three provisional patent applications filed at the USPTO (2026-06-12, pro se, micro entity): Adaptive Experimental Learning via Dynamic Diagnostic Instrument Reconfiguration (the Reincarnation engine), Reliability-Budgeted Ensemble Inference Using Psychometric Variance Decomposition (Consensus Coder), and Composition-Adjusted Benchmarking of Organizational Metrics Across Heterogeneous Data Sources. Priority dates locked; each claims a system already running in production.
  • Measurement-Meets-AI: built and test-verified a reliability module (Krippendorff's α / ICC / Cohen's κ / generalizability D-study) that treats AI systems as raters and evaluates them with a century of psychometric method — the technical spine of a thesis, and coupled book program, arguing that behavioral science directly improves AI, not only the reverse.

Selected Recent Product & Technical Work

People Analytics Toolbox · 2022–present · solo founder

One backbone for analytics, spanning 45+ live, independently-versioned microservices — psychometric diagnostics, preference modeling, privacy primitives, segmentation, statistical enrichment, compensation logic, decision forecasting, wage-compliance data, job taxonomy, and metadata-grounded SQL/Python codegen — deployed as a single Next.js application and exposed over two transports: HTTP for engineers, MCP (Model Context Protocol) for AI agents. Consumer apps vendor typed contracts; algorithms live in one place.

Technical problem solved. HR-analytics products keep re-implementing the same five primitives (anonymization, metric definitions, segmentation, surveys, decision support) and getting each one slightly wrong. The toolbox is built as shared plumbing, not another product: explicit contract versioning per spoke, k-anonymity gates as a service every team-level rollup must pass through, IRT-weighted adaptive measurement, MNL utility estimation on real conjoint designs, Wilson/t/normal CI auto-selection by data shape, and EVPI/EVSI as production-callable primitives (decision theory as a service — formal VOI in HR tooling is essentially absent commercially).

User problem solved. Adoption is spoke-by-spoke; no all-or-nothing migration. AI agents and engineering teams call the same algorithms over the same contracts. New analytical products in the portfolio sit on top of it instead of rebuilding it.

Measurement-Meets-AI · 2026–present · research + product program

A program making one case three ways: that a century of psychometric reliability theory (κ / α / ICC / generalizability theory / IRT) is the solved answer to AI's noisy-rater problem. Built a test-verified reliability module (51/51 tests) and produced the first measured result — four AI readers modeling the same document agreed at near-chance reliability on the boundaries while agreeing on a stable core (a reliable spine, unreliable edges, quantified rather than asserted). Feeds a productized reliability-scored-extraction offering, a flagship essay, and a coupled two-book program (Unreliable, trade; a practitioner manual).

Technical problem solved. AI systems that rate, score, screen, or extract are measurement instruments, and the field evaluates them with single accuracy numbers that hide the rater variance psychometrics has measured for a century. The module computes inter-rater reliability across models / runs / prompts and prices a target reliability via D-study — measurement-grade evaluation for AI, built on the discipline that already solved it for humans.

User problem solved. An organization deploying AI judgment can finally ask is it reliable, is it valid, and at what cost do we make it so — and get an answer grounded in method rather than a vendor's claim.

CanonicAI · paired engine to the Toolbox · 2022–present · solo founder

The portfolio's production line. Ingests books and primary-source archives at chapter-respecting fidelity and runs a roster of named factories — Persona, Survey, Competency, Models, Requirements, Prompt, Publishing, Business Ideas, Application Designs — producing structured behavioral artifacts that downstream products consume over a versioned REST + MCP contract. Two shells: a local AI-controlled engine that does the ingestion and extraction work, and a Vercel-hosted API host that consumers integrate against.

Technical problem solved. Cross-product behavioral measurement was incomparable because every consumer kept its own slightly-drifted definitions of competency, persona, engagement, and effect-size. The factories pin the canonical vocabulary in one place; consumers vendor typed reads against it rather than re-deriving. Cryptographic provenance (SHA-256 per source file, hash-verify-before-delete invariant) keeps the library from losing source material to a careless deletion.

User problem solved. A solo founder shipping multiple analytics products can't afford to re-extract or re-classify the same corpus once per product. The engine extracts once; every product reads the same indexed library. Dual-grade ingestion (curator picks + research bulk in one database, tag-distinguished) means one corpus serves two categorically different uses without duplication.

Namesake · namesake.baby · 2025–present · live, solo founder

A baby-naming platform built as a decision instrument, not a browse tool — guided-selection wizard, AI-enriched intelligence profiles, parallel partner brainstorming, tournament-bracket voting, themed collections, and a live baby-shower mode. Twenty years of weekly search data sit underneath.

Technical problem solved. Vendors show annual SSA rank trends; none attribute the cultural cause. Built a continuous attribution pipeline that maps name surges to specific films, shows, royal events, and other cultural moments by composing 15+ external data sources. Underneath the product: a publication-grade cultural-diffusion research apparatus (Bass diffusion, Hawkes processes, Moran's I, Granger causality, Lieberson null model, phonetic-spillover graphs).

User problem solved. Parents make a high-stakes irreversible decision based on listicles. The bracket UX forces converging on a single name rather than endless browsing; the score is deterministic and percentile-normalized so the display number is reproducible and bounded. Blank slate to shipped in 3–4 weeks.

Vela · vela.study · 2024–present · live, solo founder

A contemplative platform built around figurative art and longform editorial work — adaptive image sequencing, a five-layer written system (essays, testimony, fiction, audio, submissions), an Editorial Office for writer collaboration, and a preregistered research program running underneath. Museum-grade attribution and license discipline as a first-class feature, not a footnote.

Technical problem solved. Most "AI research" tools retrieve passages and summarize them — they don't extract patterns. Vela's synthesis pipeline runs at roughly $0.13 per research-run across multiple sub-questions with primary-source citations, at 300K+ passage scale — 10–100× cheaper than the industry norm because the discipline is statistical: extract patterns once, cite the evidence, don't re-retrieve. The Reincarnation engine reads per-reader desire on IRT-weighted axes and sequences the pool across visual rhyme and emotional register, treated as a scientific instrument under formal scrutiny (segmentation explorer, replay pipeline, instrument-validation cohort).

User problem solved. Image platforms either flatten taste into engagement metrics or hide behind gatekeepers; editorial platforms publish on a calendar rather than to readers. Vela paces per-reader — each reader's magazine begins when they arrive. Stripe membership in live mode.

Fourth & 2 / Gridiron Platform · in progress · solo founder

A fantasy-football intelligence platform built around four converging efforts: a GM workflow app (lineup, waivers, trades, draft, rankings) on a provider-agnostic adapter contract; a Python + FastAPI analytics service with PRISM rankings and the CAMS framework adapted from people analytics; a composable Insight Card system (PRISM trend, CAMS alignment, market-signal cards) surfaced both as a library at /insights and in-context across product surfaces; and a coaching-strategy simulation game layered on fantasy leagues.

Technical problem solved. Cross-portfolio framework reuse. The same Capability / Alignment / Motivation / Support model that drives Performix's diagnostic instrument adapts cleanly to football roles, schemes, usage, and offensive-line support — proof that what CAMS captures isn't HR-specific but a general theory of what produces realized performance given the conditions. The MFL adapter is the first integration against a formal provider contract; the platform is provider-agnostic by design.

User problem solved. Fantasy products are either marketplaces with shallow projections or hardcore stat tools that don't make a case. The middle — readable intelligence with a point of view — was empty. Insight Cards carry typed metadata (player, week, decision-type, confidence band) so context filtering decides when each one appears.

Penwright · 2026 · in early build

An AI-augmented authorship environment built on a fundamentally different bet than most AI-writing tools — measuring whether the writer becomes more capable over time inside the tool, not whether the output gets fluent faster. Authorship Packets replace freeform prompting (intent · structure · key ideas · relevant passages · counterpositions assembled before the AI is invoked); a Corpus Control Layer lets the writer choose which sources influence the work rather than inheriting the model's training distribution.

Technical problem solved. AI writing tools optimize for output fluency; the longer-term cost (capability erosion, voice flattening, sycophancy spirals, source attribution buried) is barely measured because the field measures what's easy to measure. The Penwright Measurement Framework — six skill dimensions, six derived indices, three measurement layers — pins the longitudinal test: better writer with Penwright than without it in six months. Anti-invention constraint at two layers (per-rendering invented_content[] warnings + Sonnet critic pass) makes the tool refuse to fabricate biographical material rather than confabulate.

User problem solved. Writers can become more capable inside an AI-augmented environment rather than be quietly substituted. Pattern-first compositional scaffolding retrieves structural moves from a curated memoir corpus without exposing source sentences (plagiarism-distance enforcement at two layers — bigram overlap + critic).

Product & Technical Capabilities

AI-Native Product Development — Retrieval-augmented synthesis pipelines at corpus scale (pgvector, embeddings, structured outputs). Multi-model orchestration (Claude Opus/Sonnet/Haiku, OpenAI embeddings, multi-provider AI Gateway). Agentic workflows with budget + quality trade-offs; structured-output prompt engineering. Modal + Supabase + Vercel deployment patterns for AI workloads. Semantic search, reranking, diachronic semantic drift analysis.

Decision Science & Quantitative Methods — Monte Carlo simulation, probabilistic modeling, scenario analysis. Value of Information (EVPI/EVSI), Kepner-Tregoe structured decision-making. Conjoint, MaxDiff, discrete choice modeling. Attrition prediction, survival analysis, compensation modeling. Time-series analysis, predictive modeling, machine learning fundamentals.

Data Architecture & Platform Engineering — Postgres + pgvector, BigQuery, Supabase, drizzle-orm, Modal (Python serverless), Vercel Fluid Compute. Shared-service architectures with typed Zod contracts + explicit version semvering + MCP transport for AI-agent consumers. Statistical-enrichment engines that auto-select confidence-interval methods by data shape; metric × segment × period combinatorial factories. Ingestion pipelines: PDF/DOCX/EPUB/audio extraction, paragraph-respecting chunking, dual-grade tagging. Row-level security, service-client patterns, migration automation, cryptographic content provenance.

Product Design & Management — Multi-surface architecture (player, magazine, admin, partner, API). Admin-over-CLI operational patterns (every script has an admin UI path). Progressive disclosure, curator workflows, taxonomy + content modeling. 45+ analytical microservices + paired CanonicAI engine + multiple consumer products shipped solo; ecosystem registry maintained as hub. Multi-agent engineering coordination (Claude Code + Cursor + DevPlane's completion-block protocol).

Program Leadership & Client Delivery — Founded PeopleAnalyst 2012; 25+ enterprise clients including Juniper Networks, Mars, Pfizer, Zoom, Reddit, Instabase, Articulate, Nike, Pure Storage, Cityblock Health, 10X Genomics, Atlassian, Udemy, New York Times. Stood up People Analytics from zero at Merck, PetSmart, Google. Led M&A pay structure reconciliation across 45+ countries (Juniper). Implemented Workday SaaS HRIS at Otsuka: 3 applications consolidated to 1 in under 10 months; $6M monthly payroll, $0 errors.

Career

PeopleAnalyst — Founder, Principal Consultant & Product Builder · Mar 2012 – present · Pittsburgh, PA (previously Dallas, Austin, Springfield MO, Portsmouth VA)

  • Founded the first niche people-analytics consulting practice in the US; served 25+ enterprise clients across pharmaceutical, tech, retail, healthcare, media.
  • Pivoted to AI-native product development in 2022; shipped the People Analytics Toolbox (46 live analytical microservices over HTTP + MCP) and the paired CanonicAI engine, plus consumer products in adjacent domains (Vela, Penwright, Namesake, Fourth & 2, Performix, DevPlane, Bicycle Guide).
  • Published People Analytics For Dummies (Wiley, March 2019); built 22K LinkedIn audience through 97+ published articles.

The New York Times — Consultant, Workforce Analytics & Compensation · Aug 2025 – 2026 (contract, completed) · Pittsburgh, PA

  • Designed a workforce intelligence model integrating merit, incentive, RSU, and discretionary-award data for executive program-design and annual-compensation planning.
  • Applied Monte Carlo simulation to compensation scenarios — advanced planning accuracy from "could miss financial targets" to reliable predictions across outcome ranges.
  • Participated in implementation of One Model cloud-based people data warehouse.

OneModel · OpenComp · AnyComp.AI — Product Strategist / Product Manager / Principal Engineer (concurrent, multi-year) · 2015 – present

  • Multi-year stints at people-data / compensation-analytics startups designing workforce intelligence and compensation decision products.
  • AnyComp.AI is own AI-native compensation-decision product line; extension of the PeopleAnalyst Platform.

Otsuka Pharmaceutical — Manager, HR Information Systems & People Analytics · Jan 2010 – Feb 2012 · Princeton, NJ

  • Designed HR technology architecture enabling a Japanese conglomerate to launch its first North American pharmaceutical business.
  • Led Workday SaaS HRIS implementation: 3 applications combined into 1 in under 10 months; outsourced Benefits + Payroll; live with $6M+ monthly payroll, $0 errors.
  • Scaled Otsuka's geographic footprint 10× with constant support headcount.

AstraZeneca — Field Researcher, Field Sales · Oct 2008 – Jan 2010 · Phoenix, AZ

  • Pioneered data-driven sales-training evaluation: proved statistically that removing salespeople from the field for sales-data training produced better subsequent performance; led to program expansion.
  • Designed dashboards combining previously isolated performance and sales data for executive talent-management discussions.

Google — Program Manager, People Analytics · Jul 2006 – Aug 2008 · Mountain View, CA

  • Pioneered People Analytics in Benefits during 7K → 21K growth and Fortune #1 Best Company to Work For era. Insights influenced multi-billion-dollar annual benefits + engagement investment.
  • Designed Google's first attrition prediction model.
  • Designed Google's first professional global employee survey — precursor to Googlegeist.
  • Partnered with BI team on Google's first HR data and reporting environment.

PetSmart — Sr. Quantitative Analyst, Talent Management · Mar 2005 – Jul 2006 · Phoenix, AZ

  • Pioneered the Quantitative Analyst role in HR at PetSmart (20K employees, $40M+ HR investment).
  • First to connect brand-specific employee measurements to business outcomes in specialty retail: analyzed how store-associate knowledge / engagement / attrition correlated with customer satisfaction and sales.
  • Identified store jobs disproportionately critical to sales; redirected performance management, compensation, and training investments accordingly.
  • Instrumental in PetSmart's "Mart to Smart" transformation.

Merck — Sr. HR Decision Support Analyst, Workforce Planning & Analytics · Jul 2001 – Mar 2005 · Whitehouse Station, NJ

  • Contributed to the development of one of the first corporate HR Analytics functions in the world, for a 65K-employee global workforce and $4B HR investment during Merck's pioneering-early era of data-driven HR.
  • First project analyzed e-learning vs. "fly-them-in" training; consensus to standardize on eLearning saved millions — the original instance of what later formalized as Rapid Collaborative Insight (RCI).
  • Analytical work across Organization Learning, Talent Management, Workforce Strategy, HR Digitization, Leadership, Engagement, Diversity.

Methodologies & Frameworks (Originated)

Placeholder one-liners — these deserve fuller introductions than this section currently allows. Earmarked for a follow-up pass before final.

  • Rapid Collaborative Insight (RCI) — combines collaboration with data-driven analysis to guide complex organizations toward optimal decisions faster than either approach alone.
  • Net Activated Value (NAV) — unifying leadership metric tying quarterly human-capital measurement to dollar outcomes.
  • Three A's Framework — Attraction, Activation, Attrition as the lifecycle measurement spine for any people-analytics function.
  • Lean People Analytics — applies leading-edge analytics practices to resource-constrained, rapidly-growing companies; extends Lean methodology into HR.
  • Full-Stack People Analytics Systems — end-to-end pipeline from data collection through storytelling, treating the analytics function as a software stack rather than a reporting team.
  • Quantitative Model of Human Resources — mathematical framework for HR resource allocation and investment.
  • CAMS — (applied in Fourth & 2 analytics API; full definition pending follow-up pass)

Publications & Public Work

  • People Analytics For Dummies — Wiley, March 2019. First mainstream book on people analytics.
  • 97+ published articles (LinkedIn, 22K followers) on Lean People Analytics, Talent Acquisition Analytics, the Five Models of People Analytics, HR Metrics, Employee Engagement, People Analytics history & philosophy.
  • People Analytics Community on LinkedIn — group moderator, thousands of members.
  • Conference speaking + podcast appearances.

Education

  • University of Minnesota — Carlson School of Management. M.A. Human Resources & Industrial Relations, 1999–2001. GPA 3.8. Student exchange: Stockholm School of Economics, Jan–Jun 2001.
  • Northern Arizona University — B.S. Sociology & Psychology, 1994–1998.

Certifications

  • SPHR (Senior Professional Human Resources) — SHRM, 2009
  • CCP (Certified Compensation Professional) — WorldatWork
  • CMS (Compensation Management Specialist) — CEBS / Wharton, 2009
  • Targeted Selection — DDI, 2009
  • MBTI Certified — Otto Kroeger & Associates, 2003
  • Corporate Online Instruction — NYU, 2002

Technology & Platforms

Languages. TypeScript (primary — every portfolio product), Python (FastAPI services, Modal serverless, extraction pipelines), SQL (Postgres, BigQuery), R, Bash; Markdown / MDX for content authoring; LaTeX (XeLaTeX via the Principia book-build orchestrator).

Software Development Tools. GitHub + GitHub Actions for source control and CI. Cursor and Claude Code for AI-augmented engineering; V0 for design-first scaffolding; Replit for hosted prototyping. Vercel (Fluid Compute, Next.js platform, AI Gateway) and Modal (Python GPU + serverless) for deploy. Supabase for managed Postgres + pgvector + Row-Level Security + Storage + Auth. Drizzle ORM, Zod for typed contracts, Playwright for end-to-end testing. Stripe for payments. Resend for transactional email. Clerk for auth where the surface needs it. MCP (Model Context Protocol) for AI-agent service integration; DevPlane for cross-tool kanban and completion-block protocol enforcement.

APIs & Data Sources Worked With. Range demonstrated across categories:

  • Museum. ARTIC (Art Institute of Chicago), the Met, BnF (Bibliothèque nationale de France), Smithsonian, Europeana, Rijksmuseum.
  • Scholarly & primary-source. OpenAlex, CrossRef, Semantic Scholar, CCEL (ANF/NPNF), Project Gutenberg, Internet Archive.
  • Cultural & naming. SSA (145+ years of names data), OMDb, TMDb, WhoisFreaks.
  • Government & geographic. US Census, TIGER 2024 county + CBSA polygons via PostGIS.
  • AI providers. Anthropic (Opus / Sonnet / Haiku), OpenAI (chat + embeddings), Google Gemini, FAL (image generation), FLUX / SDXL — orchestrated via the Vercel AI Gateway for multi-model fallback and observability.
  • Workforce & comp. Workday SOAP HRIS sync, OneModel adapter, CompAnalyst, Radford, Mercer, WTW, LinkedIn Talent Insights, Lightcast (EMSI / Burning Glass).
  • Fantasy sports. MFL (MyFantasyLeague) adapter on a formal provider-contract pattern.
  • Self-published. People Analytics Toolbox v1 REST + MCP, CanonicAI v1.2.0 REST + MCP, Principia REST + MCP (canonical-priors contract).

Data & Analytics. Postgres + pgvector, BigQuery, drizzle-orm; embedding retrieval at 300K+ passage scale; SPSS, Alteryx, Excel, Google Sheets in legacy analytical contexts.

Workforce Intelligence Platforms. Visier, One Model, ZeroedIn, LinkedIn Talent Insights, Lightcast (EMSI / Burning Glass), CompAnalyst, Radford, Mercer, WTW.

Dashboards & Visualization. Tableau, Power BI, Visier, One Model, MicroStrategy, Cognos.

HR Systems. Workday, Oracle HR, PeopleSoft, Greenhouse, Syndio, Payscale, MarketPay.

Survey & Research. Qualtrics, Culture Amp, Glint, SurveyMonkey, MaxDiff, Conjoint — plus own survey platform (Survey Respondent, 15 question types).

Current Focus (2026)

Simultaneously: (a) building out the AI-native People Analytics Platform ecosystem; (b) shipping Vela + Penwright + Bicycle Guide in adjacent domains; (c) applying to PhD programs at Carnegie Mellon and University of Pittsburgh; (d) seeking founder-track placement OR senior product leadership in applied-analytics startups, AI-native platform companies, and people-analytics platforms extending into AI.

Open to: founder-track programs (SPC, EF, On Deck, YC Visiting Group); Senior / Principal / Staff PM at applied-analytics SaaS; VP Product at legacy ERP companies modernizing with analytics + AI; Head of Product / AI at early-stage applied-analytics startups; research-leaning product roles at AI-native platform companies.

Not currently seeking: HR Business Partner roles, HR support function roles, traditional HRIS administrator roles, individual-contributor analyst roles.