2024–nowBuild
pebl
Planned, architected and built an end-to-end encrypted platform with an agentic AI production team: passkeys, multi-tenant orgs, billing and an audit ledger.
COACHPrincipalCoachExecutive◉ Austin, TX
Adrian L.✓
"The win is not a smarter agent. It is the command center your team can actually steer from."
Get matched with the right coach
Thirty years in software architecture. Co-founder and CTO of pebl, built solo with an agentic AI production team. Creator of the CMDR multi-agent platform.
Request a coach →Tell us the problem. We match you with the coach who fits it within 48 hours. Independent, not affiliated with Anthropic.
YesVirtual 1:1
coaching
coaching
7Engagements
2Years with Claude
1Workshop hubs
·CCCA credentialed
About
A coach who respects your time.
I have spent thirty years in software architecture and delivery, from IT consulting in the nineties to co-leading a 400-person digital agency as COO and CIO, where engineering execution grew with the company from $1M to over $50M in revenue.
Today I am co-founder and CTO of pebl, an end-to-end encrypted platform I planned, architected and built with an agentic AI production team: roughly 145,000 lines of production code across an API, a web client and a native iOS foundation. I also designed CMDR, a multi-agent orchestration platform where specialized AI engineering agents work under an architect's governance, with an activity ledger that shows every step.
I coach engineering leaders and founders on the AI-native software lifecycle: how to plan, review and ship when agents write most of the code, and how a senior engineer stays accountable for every merge.
Best fit
Team1-100 people
StageSeed to growth
IndustriesB2B SaaS, regulated and security-sensitive products, devtools
RolesFounders, CTOs, architects, engineering leads
Not forTeams looking for prompt templates rather than a delivery practice
Methodology
Three principles on every engagement.
i.
Map the actual workflow
We write down how your team ships today, by hand, in detail. Most teams cannot do this. That is the first finding.
ii.
Build the command surface
Agents are only as useful as the surface you steer them from. We design that surface and the review gates first.
iii.
Ship one loop end to end
One real workflow, closed loop, in production, with a human reading every risky diff. Then we scale it.
Specialties
Toolkit
Tools Adrian reaches for.
Primary model
Claude Opus 5.5 + Sonnet 5.5
Opus for architecture and review, Sonnet for throughput.
Editor
Claude Code
Default working environment for most engagements.
Orchestration
CMDR
The multi-agent platform I built for running Claude across real engineering work.
Want a coach like this?
Tell us the bottleneck in a short request. A person on our team reads it and matches you with the coach who fits it, over video or in person, usually within 48 hours.