We can't find the internet
Attempting to reconnect
Something went wrong!
Attempting to reconnect
Applied AI Engineer
Ayo Mosanya
I help organizations get more value out of their data — cleaner pipelines, sharper reporting, and systems people actually rely on. Nine years across analytics and data engineering, the last two spent teaching AI tools to do the heavy lifting.
Data pipelines are plumbing: invisible when they work, expensive when they don't. My job is making sure yours stay invisible.
Experience
Software Engineer, AI & Automation
Mar 2022 – Present →
Charles Schwab
- Build and maintain the Python pipeline that extracts, validates, and reports on enterprise risk data
- Use GitHub Copilot and frontier language models throughout the analytics workflow
- Author procedural standards for the risk analytics function and run the adherence and oversight program for how data, reporting, and AI-driven processes comply with them
- Deliver reporting datasets for Federal Reserve examination cycles
- Build and maintain Tableau & Power BI dashboards for enterprise risk reporting, with training and documentation for risk managers and issue owners
- Mentor analysts toward promotion and sit on technical interview panels
Data Analyst
Jun 2021 – Jan 2022 →
2ndWatch
- Cloud data warehouse optimization analyses on Oracle DB for Fortune 500 enterprise clients
- Developed a unified Excel modeling framework integrated with Oracle DB SQL queries
- Redesigned cost analysis processes to improve data reliability and streamline client reporting
M&A Analyst
Jul 2019 – Jun 2021 →
Salesforce
- Automated executive reporting pipelines using Tableau and Google Sheets
- Led analytics integration for technology acquisitions, consolidating data from disparate sources
- Partnered with Corporate Development, IT, and Operations to align technology integration strategies
Technical Advisor
Jun 2017 – Jul 2019 →
Apple Inc.
- Provided technical support and troubleshooting for Apple hardware and software products
- Analyzed support interaction data to identify product improvement opportunities
Education, skills, and full role details on the About page.
What I do
Data pipeline engineering
Think plumbing, not code on a page. I design the systems that move your data from where it's messy to where it's useful — and make sure they don't leak.
Reporting & analytics
The right number, in front of the right person, at the right time — without someone pulling an all-nighter in a spreadsheet to make it happen.
AI-augmented workflows
AI is the power tool, not the craftsman. I use it to compress weeks of manual work into hours — without cutting corners on accuracy.
Data quality & governance
Data you can bet on. I build the guardrails so nobody has to double-check the numbers before a board meeting.
Data pipeline engineering
End-to-end ETL design using Python, SQL, DuckDB, and Polars. I cut through manual processing, reduce errors, and build pipelines that run without babysitting.
Reporting & analytics
Dashboards and automated reports that give decision-makers the right numbers at the right time — without a data analyst in the loop every time.
AI-augmented workflows
LLM-assisted development with frontier models and GitHub Copilot that compresses development cycles and removes bottlenecks in data work — without trading speed for correctness.
Data quality & governance
Frameworks that make your data trustworthy — lineage, quality controls, and governance practices that give teams confidence in what they're working with.
Why it matters to a business
Faster decisions
Leadership gets the number when they need it, not three weeks later.
Measured by: report and dashboard turnaround time. My last pipeline rebuild cut processing time 85%.
Lower operating cost
Automation absorbs the reporting work that would otherwise need another analyst or two on the team.
Measured by: headcount or contractor hours avoided. One person, AI-assisted, producing team-level output.
Lower regulatory risk
Clean data and clean audit trails mean fewer surprises when a regulator or an auditor comes looking.
Measured by: exam findings and error rate. Most recent Federal Reserve exam closed clean; error rate held under 1%.
Faster ramp on new systems
New stack, new vendor, or a newly acquired company's tech — I get productive in it fast instead of stalling a project for months.
Measured by: time-to-production on unfamiliar tools. Zero Elixir knowledge to a deployed app in two weeks.
Things I've built
kineticform
A coach's eye, running on a phone camera.
A workout-tracking and coaching platform: log workouts, connect with a coach, build and share custom programs, and measure height and exercise form straight from your phone camera — then share that analysis with your coach. Integrates with Apple Fitness.
I call this machine learning, not AI, on purpose. Estimating height or scoring exercise form from a camera is approximation — the same category Apple's own frameworks file under machine learning rather than the generative AI everyone's currently branding everything as. I'd rather be precise about the label than chase the trend.
ayomos.com
The fastest way to prove you can learn something new is to ship it in public.
This site: built from zero Elixir knowledge to a deployed, production Phoenix LiveView app in two weeks, with AI doing a lot of the pair programming. A coding-speed story, not a shipped-ML one — that's kineticform, above.
population insights
A census you can spin, not just download.
A dashboard analyzing 2025 population projections for 233 countries and territories: real Pearson-correlation analysis surfaces relationships like fertility rate versus median age, rendered through D3.js charts and a 3D globe you can rotate, zoom, and click through country by country. Originally a flat Canvas2D globe with a hand-rolled quadtree for hit-testing; rebuilt on WebGL (three.js) for full-detail geometry everywhere and GPU-accelerated hit-testing instead.
Six data stories, an interactive globe, and a data page where the dataset can be refreshed in the browser. Works best on a wider screen; the full site link opens it on its own.
Most of what I build day to day lives behind a corporate firewall — things like AI agents that draft and validate SQL against enterprise data with minimal hand-holding, context systems that let one person produce team-level output, and tools that turn unstructured documents into structured, queryable data. I can't show you the code. I can tell you how it works — ask me.
Recent writing
All posts →Questions people actually ask
What do you actually do, in one sentence?
I turn messy, high-stakes data into pipelines and systems people trust — and lately, I use AI to get there faster.
What's your stack?
Python, SQL, DuckDB, and Polars for the data side. Frontier models — via GitHub Copilot and direct API access — for the AI-assisted side, plus some hands-on time self-hosting open-source models. Tableau for anything that needs a dashboard. I also just taught myself Elixir and Phoenix LiveView to build this site.
Have you actually shipped anything with AI, or is that a buzzword on this site?
Building this site with AI pair programming is a coding-speed story, not a shipped-ML one — that's kineticform, which measures height and reads workout form from a phone camera. I call that machine learning, not AI, on purpose: it's the same kind of on-device approximation Apple's own frameworks file under ML rather than the generative AI wave everyone's branding everything as.
Do you work with regulated or high-stakes industries?
Yes. I work on the AI & Automation team within Corporate Risk Management at a major financial institution, including Federal Reserve examination readiness — the most recent exam closed clean. I'm comfortable in compliance-heavy environments where the data has to be right the first time.
How do I get in touch?
Use the contact form or email ayo@ayomos.com. I read every message and respond within a few business days.
Background
9 years
Data engineering & analytics
Charles Schwab · Salesforce · Apple
Where the experience was built
Python · SQL · DuckDB
Polars · Tableau · Oracle DB · Frontier models