Your next dev team is a fleet of AI agents.
Your engineers already use AI. So does everyone you compete with. The fastest teams now have agents writing, testing, and reviewing code in parallel, with a couple of engineers steering. They ship more with fewer people, and the code still holds up.
30-minute call. You leave with a plan either way.
Teams I've Worked With
The problem
Sound familiar?
Falling behind
Everyone got faster at the same time.
Your competitors bought the same AI tools you did. Shipping faster than last year doesn't count for much when they are too.
Speed
AI speeds up a task here and there.
One engineer saves an afternoon on a migration. Someone else gets tests written quicker. The roadmap as a whole still moves at the old pace.
Cost
The usual fix is more engineers.
More people means more salaries, more onboarding, and more meetings. You'd rather get there with the team you have.
Quality
It all still has to work.
Faster is no use if every release comes with a fresh batch of bugs and code nobody on the team really understands.
The tools are there. What’s missing is the system around them.
The fix
What a fleet of agents looks like
Several agents work on your codebase at once, each with a clear task, your conventions, and enough context to do it well. Tests and review catch problems before anything merges. An engineer decides what gets built and signs off on what ships.
AI as a tool
- Every engineer prompts their own way
- One task at a time
- AI code gets reviewed when someone has time
- More output means hiring more people
- AI can reach more than it should
AI as a team
- One shared setup the whole team uses
- Several agents working in parallel
- Tests and review on every change before it merges
- More output from the team you already have
- Agents only get the access they need
How it works
How I build your fleet
Map the work
We start with your roadmap and your codebase. I work out which jobs agents can take on, what they need to know, and where a person has to stay in the loop.
Set up the agents
Agents for the work that eats your team's week: features, tests, refactors, migrations. Each one gets your conventions, the right context, and one clear job.
Build in the checks
Tests, review, and CI gates catch bad output before it merges. Agents propose changes. An engineer decides what ships.
Keep it safe
Agents only get the access they need, secrets stay out of prompts, and you can see what you're spending on tokens before the invoice arrives.
Then we ship real features with it, side by side with your team, until they can run it without me.
Recent work
Products and AI systems I've shipped
Teammates
Built the browser automation service behind an AI agent platform, a full abuse prevention system, and a credential-handling agent skill that lets agents log into third-party sites without exposing credentials to the LLM.
ThinkerNotes
My own AI note-taking SaaS, built and run solo. Semantic search over a personal knowledge base with embeddings and pgvector, with resilient AI integrations that never block the user.
Eros Platform
AI content platform with a provider-agnostic LLM layer (OpenAI, Anthropic, Gemini, Ollama) and pgvector semantic search, built for an in-house marketing team.
Testimonials
What founders and CTOs say
As comfortable refactoring legacy systems as he is with the newest AI-driven development practices.
Cezar brings more than two decades of deep Ruby and software engineering experience, pairing hard-won technical mastery with an instinct for where the craft is headed. He’s as comfortable refactoring legacy systems as he is experimenting with the newest AI-driven development practices, and he approaches both with curiosity and rigor. Despite being 10 time zones apart, he made collaboration effortless by being flexible, always prepared, and always finding ways to make our overlapping hours count.
Incredibly quick to grasp the opportunity at hand.
I consider myself lucky to have been able to work with Cezar for the year that we did. We knew he had a deep understanding of the technical stack but we also found that he is incredibly quick to grasp the opportunity at hand, clearly articulate the pros and cons of various solutions and work with various stakeholders on a successful implementation.
I keep asking myself, 'Would Cezar approve this?'
I had the pleasure of working with Cezar over the years on multiple projects. He was always dependable, on time, and knew what had to be done without micromanagement. What I loved even more about working with Cezar was the ability to improve my coding skills while reading his code, which, to this day, is easy to read and maintain.
I actually keep asking myself on multiple occasions when writing new code, "Would Cezar approve this?"
It was a pleasure working with him, and I'm sure to reach out to him on new projects.
Working with me
Start with one week. Keep going if it's working.
No transformation project. You see it working on your own code before you commit to anything bigger.
One-Week Build Sprint
Your first agents, shipping a real feature in 5 business days.
We pick one real feature, prototype, or stubborn problem from your backlog. I set up agents for your codebase and ship it with them, tests included, in five business days. Your team keeps the setup and gets a walkthrough of how it was done.
- The feature, merged or ready to merge
- An agent setup for your repo: instructions, conventions, review gates
- Small, tested changes your team can actually review
- A short write-up of what to set up next
- A walkthrough call with you and your team
Money back if it doesn’t ship
We agree on the feature and what “done” means before we start. If it isn’t working and ready to merge by day five, you get a full refund.
Ongoing
Then, as part of your team
Part-time product engineer
1 to 3 days a week
from $6,000/mo
I run and grow the fleet with your team: new agents, better checks, more of the roadmap handed off. And I pair with your engineers until they can run it without me.
Full-time product engineer
Full-time
from $18,000/mo
The same, every day. I join your standups, Slack, and code reviews like anyone else on the team, without the recruiting fees, notice period, or lock-in.
Prefer a review first?
Independent Second Opinion
Already using AI across the team? I look at your codebase and how your team works with agents, and tell you in plain English what's working, what's risky, and what to set up next.
- Code health, including what AI has quietly made worse
- How your team uses agents today, and where time is leaking
- Security, performance, and cost risks, in priority order
- A walkthrough call with you and your team
How to get started
From first call to shipped work
Book a call
30 minutes. We look at what you're building and how your team uses AI today. You leave with a plan either way.
Run a build sprint
One week, fixed price, money back if it doesn't ship.
Keep going if it's working
Part-time or full-time, month to month.
About
Hi, I'm Cezar
I've been shipping production software for 18 years, most of it in Ruby on Rails, including seven at RepairPal, where I grew from one of the first engineers to tech lead as the platform scaled to 3M+ monthly users.
I build with agents every day. I write the specs, set up the agents, run several at once, and review what they produce. All those years of watching codebases grow, or fall apart, are how I know what to hand to an agent and what to keep for a person.
I also build AI into products: agent platforms, RAG and semantic search, provider-agnostic LLM layers. I've trained over 160 developers, written books and courses, and I speak at conferences.
- Shipping production software since 2008, most of it Ruby on Rails
- Tech lead at RepairPal, scaled to 3M+ monthly users
- Agent platform and AI systems shipped for Teammates.work and Institute of OM
- Builds with a team of AI coding agents every day
- Trained 160+ developers, speaker at Rails Global Summit and RubyConf
- Years shipping software
- 18
- Developers trained
- 160+
- Monthly users at RepairPal
- 3M+
FAQ
Frequently asked questions
We already use AI. Why do we need this?
Most teams use AI one engineer and one prompt at a time. That helps, but it doesn't add up to much across a roadmap. Getting agents to work in parallel, on your conventions, with checks that catch bad code, takes deliberate setup. That's the part I do.
What does a fleet of agents actually mean?
Several coding agents working at the same time, each on a small, clear task, each set up with your conventions and context. I break the work down, hand it out, and review everything before it merges.
Isn't AI-written code full of bugs?
It can be when nobody's checking. Every change goes through tests and review before it merges, and the agents follow your conventions, so the code stays something your team is happy to own.
Will this replace our engineers?
No. The point is to get more out of the team you have. Your engineers end up steering agents instead of typing every line, and the setup stays with them after I leave.
What if the sprint doesn't work out?
You get your money back. Before we start, we agree on the feature and what done means. If it isn't working and ready to merge by the end of day five, I refund the full sprint fee.
Which tech stacks do you work with?
Ruby on Rails is home, with production experience going back to 2008, and its conventions make it a great fit for building with agents. With agents doing much of the typing, the language matters less than it used to. What matters is knowing what good software looks like, and that carries across stacks. Tell me what your product runs on and I'll tell you honestly if I'm a fit.
Why not just hire someone full-time?
You can, but a senior hire usually takes two months or more to find, costs recruiting fees, and is hard to undo if it doesn't work out. I can start in days, and there's no lock-in. You keep me because the work is good, not because it's hard to let me go.
Can we interview you first?
We'll talk properly on the intro call. But most clients find that a one-week build sprint tells them far more than any interview could: you see how I work, on your real product, before committing to anything longer. If it's not a fit, we stop there.
Do you work with US companies?
Yes. I work as a business, so you get a simple invoice, no payroll or hiring paperwork. I'm based in Romania and overlap with US mornings and all of the European workday.
Do you need access to our production data?
Usually not. I start with your code, logs, and config, and we agree up front on exactly what I and my agents can access. Happy to sign an NDA.
Let's build your fleet.
30 minutes. Tell me what you're building and how your team uses AI today. You'll leave with a plan, whether we work together or not.
MAP OUT YOUR AGENT FLEET