From Engineer to Leader: Managing 40 AI Researchers

Going from IC to leading 40 people at Cruise was brutal. Here's what I wish I knew.

The Transition Nobody Prepares You For

At Uber Eats, I was a founding engineer. I wrote code. I shipped features. I knew exactly how valuable I was—just count the commits.

Then at Cruise, I became a leader. My job became making other people successful. That required a completely different skill set.

Lesson 1: Your Job Is Context, Not Code

The hardest part was accepting that writing less code made me more valuable.

What I learned:

• Engineering managers multiply impact through others
• Your job is to give your team the context to make good decisions
• The best code you write is the code you prevent someone else from having to write

Lesson 2: Diverse Teams Build Better AI

Leading 40 AI researchers taught me that the best teams aren't the ones with the smartest individuals—they're the ones with the most diverse perspectives.

Our team had:

• PhDs from academia
• Self-taught engineers from startups
• Domain experts from automotive
• Product-minded engineers

Every major breakthrough came from collision of these perspectives.

Lesson 3: Process Enables Speed

As a startup engineer, I hated process. As a leader of 40 people, I learned that the right process enables speed.

What worked:

• Weekly one-on-ones: Non-negotiable time for career conversations
• Design docs: Write down decisions before committing code
• Post-mortems: Blame-free analysis of what went wrong
• Demo days: Show progress every two weeks

Lesson 4: Hiring Is Your Most Important Job

One great hire is worth 10 average ones. But finding great people is hard.

My hiring principles:

1. Optimize for learning speed: You can't predict what skills you'll need in 6 months
2. Look for builders: People who ship, regardless of title or background
3. Test for collaboration: Brilliant jerks destroy teams
4. Hire for gaps: Build a team where everyone brings something unique

Lesson 5: Autonomous Vehicles Taught Me Humility

At Cruise, we were building systems where mistakes could cost lives. That changes how you think about everything:

• Testing isn't optional: Every line of code could kill someone
• Metrics matter: 99% accuracy sounds great until you realize one percent means crashes
• Iteration has limits: You can't "move fast and break things" with 2-ton robots

What I'm Applying Now

Building FinyxAI and BoostMyCase with small teams (less than 10 people) is different from leading 40, but the principles carry over:

1. Context over control: Give people problems, not solutions
2. Diversity multiplies: Different backgrounds lead to better products
3. Process scales: The right lightweight process enables autonomy
4. Quality compounds: Ship fast, but ship right

The Meta-Skill

The biggest lesson: leadership is about leverage. Every conversation should multiply impact. Every hire should raise the bar. Every decision should compound.

That's the game.

---

Leading a technical team? I share lessons learned on X.