Halladin Labs
Illustration of Scott presenting a model architecture

Hi, I'm Scott. I lead Analytics and Data Science teams at consumer product companies, and have built AI agents that now automate 80% of analytical work. This frees teams to focus on the questions that move the business.

Scott Halladin
Product Analytics·Experimentation·Data Science·Agentic Systems·Causal Inference·Building Teams·
Selected work

Agents I've shipped

Production systems built with my team at a consumer subscription company, not side projects. Each post walks through a real problem, how we solved it, and the impact it had on the business.

How I think about analytics

I build and scale data teams with a heavy focus on product analytics and experimentation, and more recently on agentic tools that automate the repetitive parts of the job. I've built an analytics organization from scratch, led company-wide AI adoption, and shipped agentic systems that are now used in production every day.

A team's success should be measured by how many decisions it changed, not dashboards shipped or queries run. If a report isn't moving a decision, automate it and put the team on work that does.

When is an A/B test actually done?

Sometimes, a million-dollar question. Ask two analysts, and you'll often get two answers. The experiment agent makes the same call on every test, every morning.

How it works →
Fig. 1 · Control vs. Variant 1, 95% intervals Collecting
Control0.400
Variant 10.400
0.300.4250.55 conversion rate
p-value1.000
p = 1 0.05 0.001 →
Samples160
Lift+0.0%
Progress 0%

Simulated data. As samples come in, the intervals narrow and the p-value updates. Variant 1 turns green once p drops below 0.05.

Writing

Blog posts and technical write-ups on AI, analytics and experimentation.

Interactive

The AI Acceleration

An interactive timeline of every major AI moment from 1950 to today, and why each one matters. The gaps between them have shrunk from years to days.

Explore the timeline →
Draft

From 21% to 95% accuracy: what makes an AI data agent give the right answer

Anthropic's data team published how they automate most business analytics with Claude. I'd already built a self-service data agent in production. Where we landed in the same place, where I'd push back, and the three pieces I'd build first.

Responding to: How Anthropic enables self-service data analytics ↗
Outline

Graph engineering: the next upgrade is wiring, not prompting

Why the structure around an agent, what connects to what and in which order, is starting to matter more than the prompt, and what that looks like for analytics work.

Prompted by: Towards AI on graph engineering ↗

Let's talk

Email is the fastest way to reach me.

Available for consulting on analytics, experimentation and AI agents. Reach out for rates and availability.