TellezTec is an early-stage company working on what it costs to run software as a service.
Why this problem
The thesis on the front page is not abstract. I have built and run two database-as-a-service platforms, both under real cost pressure.
At Yugabyte I took a DBaaS from a proof of concept on Kubernetes to a launched product in six months, then spent three years growing it past 300 databases and 100 production customers. The last two of those years ran against continuous downsizing. By the time I left in 2025 the team was barebones and the platform was still standing — which is where the cost question stops being theoretical.
At Couchbase I inherited the same problem in a different shape: a DBaaS run with a database team’s engineering motion, which is excellent for shipping a database and poor for operating a service. Over fifteen months median project time went from 120 days to 70, and delivery accuracy from 29% to 69% — while roughly a quarter of the team left.
Both companies had strong enterprise products and real SaaS offerings. In neither case was the product the hard part. The hard part was what it cost to deliver and operate the service around it.
Before that, I worked on Apcera: a container management system in the same territory as Kubernetes, running customer SaaS across AWS, GCE and Azure at around 100 VMs and 1,500 containers, with the whole organisation — engineers and managers alike — carrying the on-call rotation. Vendor sprawl and headcount are not something I read about.
At Embrane, later acquired by Cisco, I designed the distributed architecture of the virtual network appliance, kernel and userspace. We built too early many of the virtual network components that are now on every SaaS product. That company never even envisioned deploying as a SaaS, and yet today I see Palo Alto Networks, Tailscale and many others successfully and profitably provide those services in the market.
I believe there are now opportunities to run SaaS companies and enterprises in a more lean way, without sacrificing correctness, security, feature richness, availability and performance.
Background
Twenty-five years in software, engineer through senior director. At Oracle a small team and I wrote the CRS, the cluster framework underneath RAC. Later I ran a team of 25 across three countries.
In all I have built production code in C, C++, Go, Java, Python and Bash, and have been running distributed teams since 2007.