About
David Smith
I have been building at the boundary between business, software and data for decades, and I have worked through hundreds of companies' data and data stacks in that time, many of them during due diligence or portfolio work at a venture fund. Enough of them break in the same ways that I can usually recognize the shape of a problem before the person describing it has finished. That pattern recognition is what helps me both work with clients on big-picture strategy and dive into the data to find the "devil in the details."
Anyone can point an LLM at a system and ask what's wrong. Knowing which of its answers survive contact with the people who run the business on it is the job.
The rest is people, and it is where this work usually fails. Nobody wants to be told their numbers are wrong, and the answers only hold when the people who own the systems are the ones deciding what a record means. I learn a new business fast because that part is the fun of it, and I would rather build something smarter than something anyone could have built.
What I bring
Range, and not only across industries.
- Across domains. Enterprise software, startups, data science, and now production AI, learning each domain fast and pushing until it works.
- Across the stack. Enterprise-grade (Databricks, Snowflake, BigQuery) to scrappy startups (Postgres, Metabase), right-sized to your problem and budget.
- Across hundreds of companies. Six years inside an international venture firm's portfolio, hands-on across fintech, edtech, pharma, and more, so I tell fast whether a problem is structural or just process.
- Across strategy and build. I am the rare person who sets the C-level strategy, leads the team that executes it, and builds the highest-leverage pieces myself. An engineering degree, an MBA, and decades building let me hold the business case and the technical detail in one conversation, so nothing is lost between the strategy and the delivery.
Fractionalytics, LLC
Owner and Fractional Chief Data Officer. Coral Gables, FL. Jul 2024 to present.
Independent data strategy, foundations, remediation and customer analytics for operating companies, and data diligence for the people buying them. Selected engagements, anonymized:
Consumer software platform, subscription and continuing education (Jul 2024 to Jun 2025). The company had bought a marketing automation platform and could not switch it on, because its production database and the platform disagreed about more than 300K customer records. Nobody had audited it because nobody had needed to. Ran the reconciliation as a weekly cross-functional loop with the CMO and the product and engineering teams, my analysis flagging what did not agree and their judgment deciding what each record actually was. Six months part time took it from 300K-plus to under 5K and they flipped the switch. The work landed on a Segment CDP implementation feeding Snowflake as the new source of truth, so marketing could build campaigns without engineering in the path. The full account is in The Unsexy Data Work That Actually Matters.
Payments platform, Latin America (Oct 2024 to Oct 2025). Finance and growth calculated profit per transaction differently, millions apart on a single cost line, and a new product showed a profit in one report and a loss in the other. Reconciled the definition first, then built the marketing science on numbers both sides accepted: a Google Meridian media mix model and an XGBoost multi-touch attribution model, with graph-based canonical IDs stitching Adjust, Amplitude and CleverTap into one customer journey in Databricks. Customer acquisition cost fell 50 percent, registrations rose 20 percent, and return on ad spend moved from under 1.0 into the 1.3 to 1.5 range.
Fractional staffing services firm (Sep 2024 to present). Built the company's data stack end to end: BigQuery, dbt and Looker Studio ingesting HubSpot, Harvest and QuickBooks into executive dashboards with nightly automated refresh, in sixteen production marts under CI/CD. Now drives their operating cadence, onboarding and utilization reporting.
Blue Collar AI Labs
Head of Data and AI, co-founder. Coral Gables, FL. Dec 2025 to present.
A separate AI consulting partnership, co-founded with a colleague, where I lead delivery end to end: discovery, architecture, deployment and monitoring. Blue Collar AI Labs is its own company and its client work is not Fractionalytics work.
The work below is for one client, a private-equity-backed disaster recovery services company.
- Sales conversation intelligence. An hourly pipeline moving Salesloft calls and transcripts into Azure SQL with speaker-role attribution and idempotent recovery; an LLM workflow scoring every rep call against the client's own sales methodology into structured coaching insights; and an evaluation framework that measures and improves that scoring.
- Account monitoring. GPT-4o classification over daily news and event ingestion on Azure Functions, scanning the full customer base for acquisition and churn signals. The first three months in production flagged 5.7% of monthly recurring revenue as at-risk due solely to M&A activity. What had been a reactive account-review process became a daily prompted one. Live to customer success managers since May 2026.
- Service delivery automation. An MCP server on Azure Functions wired into Microsoft 365 Copilot, SharePoint and Azure OpenAI, turning a multi-day manual expert process into a structured, human-reviewed pipeline with automated quality scoring. Covers 100-plus tabletop exercises a year.
- Salesforce audit. The company's own systems could not agree on how many customers it had. The model did most of the digging; roughly a dozen judgment calls that determined the answer were mine.
TheVentureCity
Chief Data Officer and Operating Partner. Miami, FL. Apr 2018 to Apr 2024.
Global venture fund and accelerator. Six years running data across an investment portfolio, which is where the due diligence and the portfolio-scale pattern work come from. I still advise the Investment Committee.
- Ran data diligence on prospective investments and operating support for companies already held, applying what one portfolio company's problem taught to the next. Companies often run the same systems, break them the same ways and argue about the same definitions, which is what drives the pattern recognition.
- Founded and led a data science and engineering team of seven across four countries.
- Built the modern data stack (BigQuery, Hevo, dbt) behind Growth Scanner, an automated startup health assessment used to evaluate more than 200 portfolio companies and investment opportunities.
- Published an open-source Data Pipeline Toolkit adopted by multiple startups, and taught founders on product-market fit and customer lifetime value through TechCrunch articles and internal courses.
Anda, Inc.
Manager, Business Analytics and Insights. Weston, FL. Nov 2016 to Apr 2018.
Pharmaceutical distributor, a Teva Pharmaceuticals subsidiary above $1B in revenue. Ran data science projects for the commercial side of the business.
- Automated customer-specific product recommendations into the CRM in Python, so a rep opening an account saw what that customer was likely to need instead of a catalog.
- Raised marketing promotional gross profit 12% with a seasonal product clustering model across 10K-plus drugs.
- Found roughly 5% of the freight budget in savings nobody had seen.
- Built predictive models and executive dashboards in Tableau, and led project teams across finance, pricing, operations, marketing, sales, National Accounts and procurement.
Bayview Financial, LP
Vice President, Strategic Initiatives. Coral Gables, FL. Jun 2004 to Aug 2011.
Mortgage origination and servicing. Joined as a rotational executive trainee and left having founded the software organization the origination subsidiary ran on.
- Founded, built and led a 46-person software development organization from nothing through the subsidiary's fastest years of growth.
- Delivered loan origination products that priced and sold over $2B and closed over $750M in loan volume.
- Identified process improvement and organizational redesign worth $3M a year, and repurposed existing technology to open new business lines worth $1.5M in additional annual revenue.
- Recognized as a 2007 Top 25 Tech-Savvy Lender by Mortgage Technology.
Earlier
- Ayoba Innovation / Melt Haus, owner and independent consultant, 2012 to 2016. B2C e-commerce, systems integration and analytics for a fast-growing apparel brand; co-founded a custom-designed apparel label.
- Acceller, now Bridgevine, VP Marketing Strategy, 2011 to 2012. Grew channel revenue 20% by expanding a call-center lead generation business.
- GHR Systems, product manager, 2003 to 2004.
- Ventera Corporation, manager, 1999 to 2000.
- Andersen Consulting, now Accenture, consultant, 1996 to 1999.
Systems
Where customer data lives: Salesforce, Segment, Marketo, HubSpot, Harvest, Adjust, Amplitude, CleverTap, Salesloft.
Where it mixes with everything else: Snowflake, Databricks, BigQuery, Azure SQL, dbt, Fivetran, Hevo, Evidence.dev, Looker Studio, Tableau.
Build and methods: Python, SQL, TypeScript, R. Azure Functions and Azure OpenAI, Microsoft 365 Graph and Copilot, GCP Cloud Run, Vercel, Supabase. MCP servers, multi-agent orchestration and evaluation. Entity and identity resolution, media mix modeling, multi-touch attribution, customer lifetime value, unit economics, segmentation, cohort analysis, anomaly detection.
Education
- Wharton School, University of Pennsylvania. MBA in Finance and Strategy.
- Rice University. BS in Civil and Environmental Engineering.
Interviews and talks
What Is Data Storytelling? Visual Storytelling Today, June 2024. A 49-minute conversation with Shlomi Ron of the Visual Storytelling Institute in Miami, recorded just after six years as Chief Data Officer at TheVentureCity. It covers what data storytelling actually is, how to leave an investor with one number rather than five, why trust is the constraint on analytics and on AI alike, and how I came to data late, in about 2015, out of IT consulting and software product management, because a client handed me more customer and transaction records than Excel could hold.
Elsewhere
- Email: david@fractionalytics.io