An AI employee.
Right in the chat.
I build agents that handle the conversation and get on with the work.
3,247 messages handled.AI developer. Max Sergeev.
AI agents that
do real work.
I build agents for your inbox, your phone and your codebase.
YOUR INTENT
Build an AI agent.
Context in. Work starts.INDEPENDENTLY CHECKED
CONTEXT. INTELLIGENCE. ACTION.Ready to merge.
Illustrative workflowFROM A TASK TO A WORKING PRODUCTILLUSTRATIVE WORKFLOW
Built on Kortex. Kortex for business: concept, in development. Pieces of it already run in production.
What I build.
Where it matters.
The work is already there.
I build agents that get on with it.
An AI employee.
Voice agents
A voice.
With a job to do.
A phone concierge, built with Vapi and ElevenLabs voices.
On the phone. In the workflow.AI engineering
From issue
to merged code.
An AI engineer agent that builds for a client, inside the codebase.
335 PRs merged in 102 days.Document extraction
The numbers.
Out of the PDF.
I built an ESG extraction pipeline before LLMs could do it reliably.
Documents become usable data.Kilazi Live
One call.
Everywhere with me.
Live transcription across laptop, iPhone and Apple Watch. Spanish speech, translated to English live.
A PDF report after every call.Maps and GIS
Real places.
Working software.
I built browser CAD at Pangeam, a digitiser and cadastre fleet for a Buenos Aires property site, and a zoning platform for a Canadian proptech client.
From map data to usable tools.Animations illustrate the work. Results below each animation are from real projects.
How I work
I go to sleep.
The build goes on.
I run fleets of coding agents overnight. Parallel builders, an independent tester, and reviewers. Speed needs a second pair of eyes.
A task becomes a fleet.Scroll through the process
Built. Tested. Ready to merge.
Independent checks before the work moves on.
21
Working demos in a single day.Built for prospects.
3,064
Agent sessions across the fleet.A lot of building. A lot of learning.
The foundation
It all runs
on Kortex.
Every agent I build starts with context.
One brain. Everything it needs to know.
975learnings
80projects
My agents read these before they work. Decisions, mistakes, the things I don't want to explain all over again.
1,045,056 events in Kortex. Calls, documents, notes, code and working history, in one memory.
These numbers come from the Kortex that runs my own operation. Your company gets the same foundation.
Explore KortexThe bigger idea
My AI agents
write AI agents.
I live and breathe this. My vision is an agent that learns, adds tools and expands what it can do. Give it a goal, not a task list.
Draft.
It writes.
You send.
Ask.
It checks
with you first.
Automate.
It just
does it.
Your choice, on every automation. My bet: every sustainable business will run its own AI brain. One place for its conversations and decisions. Insights straight away. An ongoing analyst. A foundation for every agent and every new hire.
Kilazi Live, the copilot, the agents, the reports: all my AI products in one. I'm dying to build one for a company.
Hire me or not: start collecting your data now. Transcribe calls, remove personal information, save the history. That's what your future brain will learn from.
Upwork Top Rated Plus
The work
speaks for itself.
100%Job Success
$400K+earned on Upwork
41jobs on Upwork
11,303hours on Upwork
Front-end engineer at Pangeam.
Bring your case.
What should your
agent do?
Tell me where the work gets stuck.
Let's build something that moves it forward.
Good questions.
Straight answers.
What can an AI developer build for my business?
I build agents for WhatsApp, phone calls, engineering and back-office work. I also build document extraction and live transcription. Bring the work you want to get off your plate; we can work out where an agent fits.
How fast is a proof of concept?
A focused proof of concept can take hours. I built 21 working demos for prospects in a single day. Your timeline depends on the scope, the data and access to your tools. A demo and a production system are different milestones.
Which models and tools do you use?
I choose today's AI models around the job, then connect them to the tools the agent needs. My work includes Vapi, ElevenLabs and Kortex. The useful question is whether the whole workflow works.
Will it work with my stack?
I start by checking your stack, data access and integration options. Then I test the smallest useful workflow. I won't promise compatibility before looking.
Who owns the code?
We agree code ownership, handover and any third-party licenses before work starts. I make those terms explicit before the build.