Gio Granato

agents over dashboards · posting what I'm building and what's breaking · SF · data + ML + AI

[Gio Granato @ggranato]

I just joined the waitlist for Tau Robotics (@taurobots): a humanoid robot that cleans your home, one hour for $30. Join with my referral link and I move up the list: https://t.co/O5b9g7uKHw

[Gio Granato @ggranato]

An engineer on our team built an AI merge advisory bot. It reads the full pipeline context - SonarQube, Dangerbot, all signals - then decides if the MR is safe to auto-approve

Documentation changes, version bumps, whitespace fixes. Stuff humans rubber-stamp anyway

The bot reviews itself. Literally installed on its own repo. Meta and beautiful

[Gio Granato @ggranato]

36 new dbt models shipped in one week across 10 teams. Revenue forecasting, verification TAT, fraud metrics, campaign analytics, data quality monitoring. That's a modern data team operating at velocity

[Gio Granato @ggranato]

We migrated our court file data pipelines using AI agents. Key learning: spend time upfront defining the project plan as a first-class deliverable. Specialized agents working in parallel on subtasks mirrors human org design and 10x'd throughput

[Gio Granato @ggranato]

We're now a multi anti-fraud product company

Our ML team built an SSN fraud signal that checks candidates against Checkr's historical records for inconsistent reuse across distinct identities

First customers already live, catching real cases

[Gio Granato @ggranato]

We just built a bot that finds stale feature flags, writes the cleanup code, opens the MR, addresses review comments, and archives the flag after a week in prod. Fully autonomous. We just review

The future is 🔥

[Gio Granato @ggranato]

A Node 16 to 20 upgrade had been on the roadmap for months

Previous estimates: 1-2 weeks of debugging cascading dependency failures

An engineer with AI-assisted debugging did it in 3 days

Code complete, green pipelines 🟢

[Gio Granato @ggranato]

Our ops team built a complex new capability as a POC without a single eng sprint

No roadmap changes. No dependency chains. Just the right tools and product leaders leaning in

That's the new playbook

[Gio Granato @ggranato]

Here's what I think most teams get wrong about AI coding tools: they optimize for model choice

But the real leverage is much more in the context you feed it --> your architecture decisions, your conventions, your hard-won engineering judgment encoded in years of code reviews

Wondering what would your company's engineering constitution look like?

[Gio Granato @ggranato]

We mined 1,865 code review comments across 535 merge requests and distilled 303 engineering principles into a constitution for our AI coding agents

Same model, same code

The difference is company-specific context

[Gio Granato @ggranato]

We call it the 🏛️ Checkr's Constitution 🏛️

Each principle is structured so agents can use it to write and review code against our actual standards and constraints

The result: agents with our Constitution catch real functional bugs that vanilla agents miss. Silent query parameter loss on redirects. localStorage crashes in restricted environments. Same model, same code; the only difference is Checkr-specific context

[Gio Granato @ggranato]

https://t.co/DvXLYPim1C

[Gio Granato @ggranato]

Building ML systems that run autonomously at scale, match billions of records, and catch problems before customers do

If that sounds like the kind of work you want, we're hiring in SF (link in the comments)

[Gio Granato @ggranato]

https://t.co/vL6YAwEkQ1 will help you bridge that gap

[Gio Granato @ggranato]

If you're targeting L5+ Data/ML roles, generic interview prep will get you to the final round

It won't get you through it

The difference is practiced live reasoning. Not more content consumption

[Gio Granato @ggranato]

Running a data team in 2026 means your analysts should be writing code, your ML engineers should understand the business, and your data engineers should be building for AI-first consumption

The walls between these roles are dissolving and that's a feature, not a bug

[Gio Granato @ggranato]

https://t.co/DvXLYPim1C

[Gio Granato @ggranato]

We're hiring a Senior Manager of Data Science on our team at Checkr to lead revenue data and SMB growth

This person will own the forecasting models and financial data products our leadership relies on, and run the data science push on our fastest-growing segment

If you've built data teams where shipping matters more than process, I'd love to talk

Link in comments