Cameron Spilker · Senior Analytics Engineer

Things I built and still run

Four live projects. Every one of them is something you can use in the next five minutes.

A data pipeline that grades its own forecasts, a newsletter written by AI agents, a card you can send someone today, and a health check for your HOA’s books. Keep scrolling.

01 / 04 · NCAA basketball, ingestion to dashboard, all in the open

Full Data Stack Lab

Every NCAA Division I men's basketball team tracked through the season, the tournament simulated 20,000 times, and a page grading how well the model's own predictions did.

Python pulls a public API into a DuckDB warehouse, dbt builds 20 models behind 118 tests, Dagster schedules the whole graph, and Evidence renders the dashboard from the same file dbt writes. The self-grading page is the part most bracket models leave out: forecasts that used only prior games are scored separately from ones that saw the future.

dbtDuckDBDagsterEvidence.devPython

Open the lab Read the source →

Two dashboards over one set of models: the Evidence pages, where you pick a team and the page answers, and the dbt Charts boards, each of them one YAML file. Start wherever the landing page points you.

lab.cameronspilker.com

02 / 04 · A weekly newsletter written by AI agents

Plainstocks

Every week, a team of AI agents manages one simulated portfolio and writes up what it changed, trying to beat the S&P 500 in public.

No real money is at stake. A multi-agent research desk holds 8 to 15 paper positions continuously and publishes every trade. Python gathers the data with no model involved; two Claude Code steps propose adds and exits; a judge step can only lower confidence, never invent a pick. Every position is marked daily against the S&P 500, and every issue stops at a human review gate before it sends.

PythonMulti-agentSupabaseGitHub Actions

Sign up for the newsletter

Free weekly email. Every trade and the running score against the S&P 500.

03 / 04 · Co-founded with my brother Ethan

Cardtacular

Make a digital greeting card in the browser and send it with a link.

Pick a template, then lay out text, photos, GIFs, and voice messages across a four-page card. Nobody has to make an account to build one or to open one. React and Supabase, deployed as a static SPA.

FounderReactSupabase

Make a card in five minutes

No account needed to make one. Text, photos, GIFs, and voice.

04 / 04 · AI financial analysis for HOA boards

HOApulse

Upload your HOA's financial statements and find out whether its books are healthy.

Drop in an income statement and a balance sheet; get parsed line items, a dashboard, and a read on reserve health, budget against actual, and delinquency. The parser tries deterministic strategies first and only falls back to a model when they fail validation, so a normal month costs compute rather than tokens.

Next.jsSupabasePDF parsingAI

Score your HOA's books

Upload a statement, get reserve health and budget variance back.

Analytics engineering

The Full Data Stack Lab is live

One repository holding every stage of an analytics stack: the extractors, the warehouse, 23 models and their tests, the orchestrator, and the two dashboards those models exist to serve. The landing page opens onto both.

dbt models: 23dbt models
tests: 154tests
simulated brackets: 20,000simulated brackets
teams tracked: 365teams tracked

Two ways in, both reading the same models: the Evidence dashboard, where you pick a team and the page answers, and the dbt Charts boards, where each page is one YAML file. The landing page says what each one is for.

Open the lab

Next rebuild in

--:--:--

Nightly at 06:00 Mountain · 0 6 * * *

Scores, box scores, betting lines and ratings for the season in progress, then the dbt graph behind them.

Tools

Things I needed, so I built them

Every one of these runs entirely in your browser. Nothing you load is uploaded anywhere, which is the only way a tool that reads a work artifact is worth using at work.

The same rule as everything else here: if it reads a file you would not hand to a stranger, it has no server to hand it to. These parse in the page and forget it when you close the tab.

dbt run results visualizer
See where a dbt run actually spent its time
Drop in a run_results.json and get the timeline, the idle threads, and the models worth fixing. Nothing leaves your browser.

All tools

Experience

Nine years of making numbers defensible.

Analytics engineering at Typeform, Apollo.io, and Gopuff, with data engineering and audit analytics before that. The through line is migrations, cost, and trust: the work that makes a dashboard, or an AI answer, worth relying on.

Full history on LinkedIn →

Oct 2024 to Present
Typeform
Senior Analytics Engineer
- Launched conversational analytics on the Omni semantic layer, giving stakeholders natural language access to governed metrics. Built reusable components into the dbt repo and run evals on the context behind them to keep answers accurate.
- Co-led a zero-downtime migration of 650+ dbt models and 600+ Looker assets, seven years of reporting and operational data, from Redshift to Snowflake.
- Led the Looker to Omni migration of 400+ dashboards and 75+ Explores, prioritized by real usage and delivered with an external team. Now the Omni admin for access, permissions, and user content.
- Built feature-level reporting end to end: consolidated eventing data, defined the key properties, and shipped the dbt models, semantic layer, and dashboards. Scaled from an MVP of 7 features to 145 in production.
- Introduced macros, DRY doc blocks, PR templates, and CI/CD with automated summaries and data diffs, cutting review cycles ~25% across a four person analytics engineering team.
dbtSnowflakeOmniSemantic layerAICI/CD
Sep 2023 to Oct 2024
Apollo.io
Senior Analytics Engineer
- Cut CI/CD runtime from 60 minutes to 6 through code diffing and smart caching, accelerating PR feedback for 10+ analytics engineers.
- Re-engineered 50+ dbt models and 150+ Looker assets after major Salesforce architecture changes, restoring accuracy across 20+ dashboards used by Sales, Finance, Customer Success, and Marketing.
- Streamlined reverse ETL syncs, reducing processed records from 6 million to 1.8 million and sync duration from 30+ hours to under 1, lowering compute cost and improving freshness.
- Led a company wide doc-a-thon that raised dbt documentation coverage from 46% to 85%, improving discoverability and onboarding.
dbtSnowflakeLookerCensusSalesforce
Mar 2022 to Sep 2023
Gopuff
Analytics Engineer
- Achieved $300K+ in annualized savings by optimizing Snowflake queries and pipelines across core analytics workloads.
- Designed centralized subject area models for the Growth and Product Analytics teams covering search, ads, impressions, marketing performance, and competitive pricing, increasing consistency across 10+ core metrics.
- Built and maintained Looker and Sigma dashboards enabling self serve analytics for hundreds of internal users, reducing ad hoc report requests ~25%.
dbtSnowflakeLookerSigma

Before that

2025Brigham Young University · Adjunct Professor, IS 515 Advanced Spreadsheets
2021 to 2022Lendio · Data Engineer
2019 to 2021American Express · Data Analytics & Innovation Analyst
2019Xerva, an Eide Bailly company · Business Intelligence Developer
2017 to 2019Brigham Young University · Associate Audit Analyst
2018KPMG · Advisory Intern

About

Who I am

I am an analytics engineer based in Utah. I work across the whole stack: pulling from raw APIs, modeling in dbt, orchestrating the runs, and shipping the dashboard people actually open. The parts I care most about are the unglamorous ones. Tests, lineage, and documentation are what make a number defensible.

I work AI-forward. I treat models as a way to move faster through the mechanical parts so more of my attention goes to the judgment calls: what to measure, what to trust, and what to throw away.

I studied information systems at BYU and taught spreadsheets there. Mostly I like building things with AI and working with data, and I am always happy to talk with people doing either. If something here is useful to you, or you want to compare notes on any of it, get in touch.

Based in: Utah
Studied: Information Systems, BYU (BS + MS)
Taught: IS 515, Advanced Spreadsheets, at BYU
Tools: dbt · Snowflake · Looker · Omni · Python