Things I learned at STELLA

In September, I left STELLA Automotive AI after just over two years. I started as Director of Quality Assurance and, in under two years, became Director of Technical Services. For those unaware, STELLA uses its AI-powered voice systems to help automotive dealers engage with customers. For me, this was a new product space, a challenging technical problem and a chance to build a team while working with a CTO I admire.

I want to reflect on the journey and the things I learned:

The Complexity of Blurred Lines

STELLA was spun-off from its sister company, Telepathy Labs (”the lab”). While most of STELLA operated as a separate entity, engineering and product were tightly coupled across both companies, sharing software systems and people.

This meant release details, feature ownership, and change requests were opaque due to the many systems and teams involved (ML, Ops, Platform teams, etc). JIRA reflected this madness, it felt like turtles all the way down. This created barriers to understanding the delivery process.

Focus on People and Process

Despite how challenging the delivery process was, I assumed responsibility for scattered work processes and held the line on product delivery. By becoming the customer at the lab—literally and figuratively the person who would take delivery of their work—we made the process more predictable and salient. With clearer release guidance, my team was able to communicate feature releases to the whole organization and drive feature adoption.

My test team often felt isolated and out of the loop, so we focused on fixing that. Daily stand-ups, one-one-ones, and skip-levels helped with communication and alignment. Introducing career ladders and dedicated coaching led to raises, promotions and personal growth. Critically, the team shifted from push to pull work, which led to greater participation and eventually more visible work across the organization.

Set up for Success: The Azure Migration

In early 2024, STELLA became a Microsoft partner, which meant migrating from AWS to Azure. The migration would start with the lab’s infrastructure and then touch internal staff at both companies, third-party vendors, and live customer deployments. I was asked to lead it. 

Initially, I tried to map all the work involved, but it was overwhelming and I was too slow. My CTO gave some good advice: start with hard deadlines, document what you know, and highlight the gaps for others to fill. By acknowledging what we didn’t yet know, we could plan around the unknowns and move forward confidently. We created shift schedules, trained contributors, and executed. The migration finished faster than expected with zero customer impact!

Re-platforming, Quickly

As I stepped into my expanded role as Director of Technical Services, we faced one of STELLA’s biggest technical undertakings: re-platforming. Our entire product was going to move off of the lab’s infrastructure and onto our own using modern Generative AI.

This was both a technical and organizational reset — the kind of project where everything touches everything else. The goal was ambitious: build a minimum viable platform, migrate customers, and ensure continuity, all under an aggressive timeline.

In addition to leading the outward-facing technical operations (sales, support, reporting, etc), and having my team test changes, I filled in the gaps wherever I could. I ran company-wide dogfooding, planned tests, and shared results with our new engineering team. I worked as a bridge between our support and engineering teams; contributed to cutover plan and handled migrations and customer updates when necessary.

Despite many challenges, we successfully transitioned to a new platform that gave STELLA full autonomy over its technology stack. It was one of the most intense and rewarding projects of my career. 

Lessons that Stuck

Looking back there are a few lessons that stick with me:

• Bias for action: Fill gaps when important work isn’t being addressed.
• Communication matters: Clarity and visibility transform perception and effectiveness.
• Coaching accelerates growth: Early feedback is invaluable. Giving feedback and coaching others (direct reports, peers) on how to improve is underrated but incredibly valuable. I was also the recipient of some very good coaching from my CTO and the VP of Engineering at the lab.
• Uncertainty is navigable: Identify gaps, name the unknowns, plan around them, and move forward confidently.

Reflections

I left STELLA feeling challenged and proud. The work pushed me to navigate new technical and organizational challenges, coordinate with strategic partners, and manage multiple customer-impacting migrations. I’m proud of the impact I had but I also carry some regrets.

One such regret is never getting a chance to tackle automating our testing. I would have enjoyed trying to get an AI system to test another AI system.

I’m sure I missed a ton of things in my reflection here, but I have one final thought. I just want to say how much I appreciated the folks I worked with at STELLA. You were definitely the best part of the job.