Product leader. Data platforms and data quality for AI systems.
Product leader who builds the infrastructure that makes data trustworthy enough for AI systems and people to act on without re-checking it. Twenty years across data platforms, semantic layers, and model measurement, turning one-off data requests into reusable infrastructure that compounds.
Independent research and building
- Published thesis on trust infrastructure for AI analytics. Twenty-three articles at thetruthlayer.substack.com, plus interactive artifacts at thetruthlayer.dev including one on annotation quality, showing how human label quality propagates into a model evaluation score and its confidence interval, grounded in a real labeling loop that lifted response accuracy 24%.
- Built the Trust Inspector, a React 19 and TypeScript prototype visualizing machine-readable trust contracts, metric lineage, and governance state, dogfooding the thesis. It enforces team permissions on the server, persists comments and decisions, and runs a Claude agent that authenticates with a scoped key and is graded by the same loop as human reviewers. Deployed and demoable, access-controlled; demo available on request.
- Shipped Overlay, a generative product on Claude designed, built, and operated solo, from first commit to public launch in nine days with abuse and spend controls live before the first model call. Launch was gated on a judge-scored eval bar rather than a date, five quality dimensions at a 4.0 target, cleared on 38 scored courses that included adversarial prompts. Model spend runs behind a fail-closed quota gate and a global daily kill switch with burn-rate alerting at 70%, and structured cost telemetry prices each generated course at 72 cents.
- Live in self-built AI tooling. A git-backed PM operating repo with an auto-generated dependency graph, custom Claude Code skills, and a six-agent build pipeline coordinated through machine-verifiable acceptance criteria. 159 tracked hours produced 276 artifacts including 7 BRDs and 8 executed contracts, a measured 4.55x against a modeled manual baseline with an explicit bias adjustment.
Roles
Senior Technical Product Manager
Single-threaded owner of Metrics Library, the semantic and data-quality layer replacing a 13-year legacy system for all of AWS Finance. Govern 1,000+ metrics and 100+ dimensions across 50 reporting cubes against a fixed decommission date. Prior AWS Finance roles spanned embedded analytics, product telemetry infrastructure, and enterprise governance for $25B contingent-worker spend.
- Built data quality into the platform itself. Every metric carries a four-tier source classification from verified to reverse-engineered, with five automated re-sign-off triggers that force revalidation when logic changes, so stale data never reaches a consumer or an agent. The governed catalog feeds board-level Audit Committee reporting.
- Turned recurring one-off requests into reusable infrastructure. Personally built a 54-script Python toolchain (11 modules, SigV4-authenticated APIs) that generates data contracts and sign-off packages from the legacy system, eliminating seven categories of engineering interrupts and 17 development tickets, and shipped Named Export precomputed snapshots to production, ending the manual data-pull tickets engineers ran for users.
- Defined the Named Export scheduling model that shipped, covering daily, weekly, monthly, custom cron, and workday-based schedules, with the job lifecycle, notifications, and UX requirements. Authored the platform's MWAA orchestration spec as its PM, covering ingestion DAG scheduling, schema, completeness, and referential-integrity validation, retries with exponential backoff, dead-letter handling, and real-time SLA monitoring. Serve as the escalation tier one hour after a missed publishing window.
- Defined outcome-based quality KPIs. Built a three-layer observability framework with per-layer error attribution across three independently owned components, and authored the platform direction document after research showed 60% of AI users hesitated because they could not validate answers against an authoritative source.
- Consolidated scattered instrumentation across nine engineering teams, reconciling differing schemas, collection patterns, granularity, and pull frequency, and built the ETL merging them into a single portfolio view feeding weekly and monthly business reviews. Owned the instrumentation gap decisions, what to add, what to retire as bloat, what to capture temporarily, and whether a backfill was worth the cost.
- Shipped vendor data consolidation infrastructure in a prior role. Architected the unified platform ingesting 12+ external sources (Adobe, LinkedIn, Merkle, Radancy) to measure ROI on $25M+ annual marketing spend, cutting 80+ hours per week of manual consolidation across teams.
- Serve two internal user groups from one platform, the nine source teams that govern data logic and the analysts who consume it. Ran metric-by-metric sign-off sessions with all nine source-team PMs to set the authoritative catalog, and drove alignment across nine organizations without positional authority through a 17-section data contract framework and 8 executed governance agreements.
- Took that contract framework into production as contracts as code. I designed the schema, drove implementation, and drove adoption and onboarding across the other engineering teams, with engineering building on that design. Each producer-consumer contract now stays current on its own, and a change that would reach a customer raises an alert in either direction, producer to consumer or consumer back to producer.
- Defined the agent-access surface on the Metrics Library API, nine production MCP tools through which agents discover, query, and act on governed metrics. Owned which tools exist, what each is for, the phasing to grow them, and the trust contracts they carry, with engineering building them.
- Built a cross-org knowledge system for 12 AWS FinTech products, treating the non-technical path as its own code so technical and non-technical staff get answers in their own terms about what each product does, how the products connect, and the blast radius of a change. It runs on live code and picks up product and relationship changes automatically, and a feedback hook sends questionable changes to named owners on each product team. It is open across AWS Finance and is being folded into the routing layer of the org's finance agent.
- Shipped an end-to-end analytics platform with an AI agent on top in a prior role. The team streamed production data through DynamoDB, Kinesis, and Redshift. I personally rebuilt the SQL data model to enforce row-level security, built the governed datasets and embedded dashboards, and drove the chat agent to an 89% first-month success rate. Adoption grew from 1,000 to 3,543 users in six months, on telemetry I built from scratch that became a required launch standard.
- Built that telemetry from scratch and made it a launch gate. Wrote instrumentation for core flows personally, designed the event schema enabling full user-journey reconstruction, and shipped automated drop-off alerting. Coverage went from 40% to near-complete, analysis time from 4 hours to 1.2, and the work surfaced a 22% abandonment point that drove a redesign lifting completion 15%.
Product Director, Data and Analytics
- Built the measurement layer over a human review and labeling loop for an AI assistant platform. Conversations labeled against an intent taxonomy, with precision and false positive tracking, fed model retraining that was A/B tested and rolled out in stages, improving response accuracy 24%.
- Built an early warning system so client problems surfaced before clients reported them. Wrote the usage queries and built the dashboard, set alert thresholds off percentile distributions rather than fixed rules, then validated alert quality against a week of real alerts and a manual review of the top 20 clients before shipping. 80% of alerts proved actionable, so it rolled to 40% of the Client Success team. In month one it surfaced 14 client problems that would otherwise have arrived as escalations, 7 of them optimization opportunities rather than defects. Later automated into client-specific recommendations.
- Shipped report filtering against a 6-month engineering estimate by sequencing it. Ranked roughly 40 standard reports and their filterable fields using actual report-run data plus interviews with customers and the support team, then released report by report instead of waiting for the full retrofit. Of about 100 customers using a report before filters, 90 kept using filters afterward. Data-filtering tickets fell from 4 a week to 1, and CSAT reached 4.4 across 120 clients.
- Owned the streaming and serving stack for product data, ingest on Kinesis with open table formats (Iceberg, in current use at AWS and in prior roles), Redshift, Spectrum, and Athena. Scaled the platform from 5 to 75 reporting datasets on AWS and accelerated deployment 26%, while standardizing governance and documentation that cut time spent on data requests 30%.
- Founded the product operations department, standing up four functions from nothing: Support Engineering, Data Insights and Analytics, Data Science, and QA, with the ticketing system, SLAs, documentation, and triage training behind them. Support resolution moved from two or three weeks to a two-to-three-day SLA and data turnaround from two weeks to three or four days. Carried on-call and personally owned high-severity client incidents.
- Grew a lead support engineer into the support manager, who then took over the full team reporting to me. Managed two product managers directly, including a support engineer promoted into the product role.
- Cut report timeouts from 10% to 2% within my product area, and consolidated refresh cadences ranging from every six hours to daily onto a steady one-hour window.
- Worked inside the compliance programs on the product side, SOC 2, GDPR across a global client base, and a FedRAMP authorization completed during my tenure with government customers onboarded on it.
Solutions Architect
- Owned the full implementation lifecycle for MSP (managed services provider) solutions: vendor management system deployment, integration development, testing, and rollout. Delivered across 5 major clients, improving workforce program efficiency 30% with a 20% faster deployment rate.
Director, Technical Services
- Led global technical services and executive KPI reporting for a customer data platform whose core product problem was PII handling, consent, and retention. Ran 24/7 global support and solutions engineering, reporting to the VP of Professional Services. Operated inside the SOC 2 program and ran operational GDPR work during initial enforcement.
Product Director, Data and Analytics
- Lead technical product manager for two teams, owning vision, strategy, and execution across all data and analytics products, monetization efforts, and partnerships.
- Relaunched a stale vendor-scorecard product that had lost adoption and had scaling limits ahead of it. Researched past and current usage, rewrote the user stories across more personas, defined executive-level success measures, pulled in engineering and UX, and aligned the new data science group on what the product could learn from. Reached 35 clients using it consistently with 50+ more expected to adopt, deployed across two different architectures, single-tenant .NET and multi-tenant Oracle.
- Owned SmartRate, a rate-benchmarking product. Worked with a data science team to turn normalized, masked contractor data into market-rate benchmarks and vendor scorecards, exposing no individual client's underlying data.
- Drove data architecture convergence after the IQNavigator acquisition, aligning two merged organizations with separate platforms, tooling, and teams, and launching the first two products onto the converged platform across 150+ clients for $10M in strategic value.
- Led a $300K+ build/buy/partner platform decision replacing a perpetual-license BI platform, across 14+ vendor evaluations with Gartner Magic Quadrant and capability gap analyses, moving to Power BI for $750K in annual savings and extending it to 150+ new clients. Completed a data-monetization analysis identifying $50M in annual revenue potential.
Director, Product Operations and Analytics
Reported to the CTO, later the COO.
- Grew the global shared services organization from 14 to 30 in three years across five roles: technical managers, configuration engineers, data analysts, QA engineers, and business analytics specialists. Built and led the manager layer beneath me, and sustained a 95% GreatPlaceToWork departmental ranking over four years. Member of the senior leadership team.
- Ran the product operations function (technical services plus configuration services) for a platform serving 150+ clients and 100K+ users, and built the requirements framework behind SmartView, data contracts across 11 analytics modules, configurable metric scoring across 22 metrics in 3 categories, and explicit grain definitions across 22 datasets.
- Shipped a usage-and-access reporting product for enterprise program administrators, adopted by 50+ clients averaging 100K users and cutting call and ticket volume 20%.
- Performed 125+ implementation reviews that held product standards and drove feature adoption, cutting rework roughly 60%.
- Built a per-product implementation configuration toolkit adopted across 45+ new client go-lives and 25 expansions, cutting average configuration time from a month to 10 days, post-go-live bug tickets 60%, and lifting feature adoption 25%. Later extended to sales sandbox and demo setups to shorten the sales cycle.
- Turned a manual technical assessment service into a self-service product, cutting one work week down to 30 minutes.
Configuration Services Manager and Lead Configuration Engineer
As manager from 2008 to 2012, reported to engineering leadership. Across the Beeline tenure, promoted two individual contributors into management, a lead configuration engineer to Configuration Services Manager and later Director of that group, and a lead analytics point of contact to Analytics Manager, and managed both as direct reports for four years.
- Founded the global shared services team and grew it from 2 to 14 in three years, covering software configuration and development, QA, solution architecture, launch management, UX, and defect resolution.
- Ran 25 onsite technical assessment sessions over two years for long-tenured major accounts, plus power-user shadow sessions, producing 500+ product and service improvements. Feature utilization across those accounts rose 24% on average, satisfaction moved from 3.6 to 4.3 CSAT, and three of the 25 were saved from non-renewal.
Selected skills
Data platform and semantic layers, data contracts and SLAs, contracts as code, scheduling and orchestration requirements (MWAA), provenance and lineage, metadata and catalog governance, third-party data ingestion, observability frameworks and instrumentation coverage, telemetry pipelines, LLM cost telemetry and spend attribution, eval gating and guardrails, streaming ingest (Kinesis, Iceberg, Redshift, Spectrum, Athena), row-level access control, SQL, Python data tooling, data quality and evaluation frameworks, model measurement and monitoring, human-in-the-loop validation, SOC 2, GDPR and FedRAMP program exposure, AI and LLM-powered products.
Education, certifications and awards
- Udacity, Nanodegree in Python and Computer Science, 2017 to 2018
- University of North Florida, Management Information Systems, 2002 to 2004
- Rutgers University, Information Systems Management, 1999 to 2001
- Harvard Business School, Leadership Principles, 2020
- Pragmatic Marketing Product Management, Foundations and Focus, PMC-II
- Blanchard, Situational Leadership II, 2016
- President's Award, Beeline, 2013. The company's most prestigious individual award, for absolute brilliance in contribution.
- Army National Guard, Sergeant (E-5), Communications, 1998 to 2004. Three hurricane activations, a one-year Force Protection deployment, and Commandant's List (top 10%) at the Primary Leadership Development Course.