I help AI startups scale faster by fixing the workflows that stall real growth.

I find where manual work is slowing you down, whether that's in your product or growth engine, and I help you automate it.

Book a Call
Sam Dultsin
↓

What I Actually Do

I work with early-stage AI startups on two things:

Growth Automation

If your outreach gets low reply rates, personalization stops at basic Apollo info, or your team needs help with scaling your campaigns:

  • I automate the research and personalization that's eating your time
  • I build systems that scale without losing the human personalization touch
  • I test and optimize your workflows so they keep improving over time

*The cold email that got you here is an example of what I build.

AI Evaluation

If your model outputs are inconsistent, you're shipping updates blind, or you don't have a systematic way to catch what's breaking:

  • I build eval frameworks that surface failure patterns automatically
  • I test edge cases systematically so you know what breaks before users do
  • I turn "works in demos" into "works reliably in production"

How I Work

When I needed a website, I didn't wait to "learn web development properly." I spent a weekend figuring out what tools existed (Cursor, Claude, GitHub, Vercel), built it, deployed it, and moved on.

That's how I approach every workflow problem:

1

Map every manual step and find where time is wasted

2

Figure out what tools, APIs, or integrations can replace the manual parts

3

Build it, test it, ship it

4

Iterate when it breaks

On College

(And Why I'm Doing This)

I'm still in college, though I'm increasingly skeptical it's the right place to be (for me at least).

The system is built to sell credentials, not capabilities. That worked when a degree was a reliable signal of competence. But AI is closing the gap between "I want to learn X" and "I can do X" faster than universities can adapt. Five years ago, building a web app with zero technical experience meant months of grinding through YT tutorials. Now it takes a few days.

The question isn't whether college still has value, it's whether the opportunity cost makes sense. I could spend the next two years optimizing for a credential that's worth less every semester, or I could spend them building cool products solving actual problems.

College earnings and debt trends

Sources: CNBC (2023), Brookings Institution (2024). Adjusted for inflation.

If an opportunity turns into something that makes college irrelevant, I'll take it. If not, I'll have spent my time solving real problems instead of theoretical ones.

What You'll Find Here

Here are a few resources that show you how I think and work:

The cold email system that got you here

A breakdown of the automated research, scoring, and personalization workflow I built to reach you.

How I'd build your eval framework

A detailed workflow for systematically testing LLM outputs, catching failure modes, and turning unreliable AI features into production-ready products.

These aren't polished portfolio pieces. They're simply working documents that show my process.

Let's Talk

If you're dealing with an AI product that's inconsistent or an outreach process that's burning time, book a 15-minute call.

We'll talk about what's not working, and I'll tell you if I can help. If I can't, I'll point you toward someone who can.