After an incredible 6.5-year run at Nickel — scaling the data team from 3 to over 40 professionals and eventually leading a 100+ person engineering department — I recently made a major career shift. I left my salaried "manager of managers" role to launch my own independent data consultancy, Unytics, and step back into a hands-on management role at Optic 2000.
When deciding what my next chapter would look like, I explored three very different paths. Here is why I chose the one I did, and what I've learned in my first few months building in the AI era.
Path 1: Staying a Manager of Managers
There is a unique thrill to high-level engineering leadership. What I loved about managing managers is the incredible leverage you have. Tiny, strategic actions can yield a dramatic impact across an entire organization. Because you are detached from the daily grind of field work, you gain a holistic view that allows you to spot connections that the people actively building simply can't see. Steering the ship from that vantage point brings huge gains in generated value.
Beyond that, you learn an immense amount by managing other managers, and exposing yourself to the pressure of leading and inspiring a large department makes you progress rapidly.
But despite all of this, I realized something important: I missed building.
Path 2: Founding a VC-Backed Startup
To scratch that itch to build, I had started creating open-source tools as a side project. My initial intention when leaving Nickel was to use these tools — and the audience they generated — as the foundation for a software company.
However, after speaking with several founders, I realized the traditional startup path wasn't for me. I didn't want to raise venture capital (which often just feels like trading one boss for a board of them), I had no desire to relocate to the US, and frankly, I don't possess that specific entrepreneurial spirit of constantly hunting for random ideas with business potential just to iterate on them. I wanted to build solutions to problems I deeply understood.
Path 3: The Independent "Hands-On" Builder (The Sweet Spot)
That's when the opportunity arose to rebuild the data platform at Optic 2000 from scratch as a freelance Head of Data Platform. It was the perfect catalyst. It was a concrete first step out of the salaried world and into entrepreneurship — allowing me to sell my time and expertise while building the foundation for Unytics.
So, I joined Optic 2000 to lead a tight-knit, highly focused team of four. Our mission? To rebuild the platform on modern technologies to empower mid-technical teams (like business analysts) with true autonomy.
For too long, data platform teams have defaulted to the "it's complicated" response, treating requests as black boxes that can take months to deliver. We are building a "yes" team — one that provides solutions today and empowers business people to answer their own needs and unlock use cases that directly generate value.
Building in the Claude Code Era
Working on this new stack has been wildly fun, largely because we are starting from scratch in the era of AI coding agents.
We are leveraging our past experiences to build a completely new platform in mere weeks. By getting my hands dirty again with a small team and relying heavily on Claude Code, we are operating at a prodigious pace — doing things I could never have executed previously.
For example, we are taking the concept of "Data Platform as a Product" a step further. We have a unique entry point (a data portal) to answer any question about both the platform and the data itself. We've consolidated the full data stack into a single monorepo, taking YAML engineering to the next level.
The workflow is a game-changer. Previously, if an analysis required data we didn't have in our data marts, you had to switch tools or submit a ticket to another team. Today, using Claude Code directly in the repo, a single conversation to analyze data can detect the interesting raw sources, load them, pipe them into marts via dbt, and execute the analysis.
(I'll be doing a deep dive into this exact monorepo/Claude Code workflow in my next post).
The Takeaway
Living through this paradigm shift — where we guide the architecture and let AI write the code — feels incredible. Partner that with the benefits of a small, hands-on team at a lower management level (easy alignment, zero unnecessary meetings, pure focus), and productivity skyrockets.
I'm excited to share more of what we are building, the modern architectures we are designing at Unytics, and the technical deep-dives of how AI is fundamentally changing data engineering. Stay tuned.