Iliya Valchanov

CEO @ Team-GPT | 1,300,000+ AI students on Udemy

Putting AI agents inside marketing agencies.

3x founder, 1.5M AI students on Udemy, ex-math competitor, finance graduate, 3x $0-$1M, boostrapped to $30M revenue.

Creator of content, software, and agents.

Co-founder & CEO

Learn, collaborate, and master ChatGPT in one shared workspace.

We are planning our biggest launch ever.

I'm in full founder mode, currently running 3 agents at a time (2 coding, 1 GTM), not counting the background ones. 

I don't have cognitive space for more even if I wanted to.

I feel like 10x-ing myself (again?). How many times can this happen within 1 year?

Will keep you posted.

undefined undefined

Personal website

Juma is a team-first AI workspace for marketing teams. 

We help teams and agencies co-create with AI, stay aligned on every project, and deliver better results together. Because the best ideas don't come from AI alone. They come from teams working together.

Juma is trusted by teams at Costa Coffee, Procore, Maersk, and Salesforce. Backed by True Ventures and Pact.


75K+ users, $1M+ ARR, AI for marketing teams

Co-Founder & CEO

1.5M+ students in the field of statistics, data science, machine learning, and deep learning.

AI Course Creator

3veta is an appointment scheduling platform with 10K+ active users.

Co-Founder

Co-founded and led 365 Data Science to 4M students and $1M ARR.

CEO @ Juma | 1,500,000+ AI students on Udemy

CEO @ Juma → AI that runs your marketing | 1.5M+ AI students in a past life

Juma was boostrapping and profitable.

But our competition wasn't. 
They had raised $3-5-10M+ and were growing by burning tons of capital.

The only way for us to exist today was to become venture-backed. 

Your market and competition are often the main drivers of VC-backed vs bootstrapped.

Trump banned Anthropic's Claude from working with US government.

It's because CEO of Anthropic refused these 2 use cases:
- Mass surveillance of US citizens
- Fully autonomous weapons

Mass surveillance of US citizens is clearly bad.

But here's why autonomous weapons are so terrifying:

Imagine 10 million drones controlled by a single person, with no chain of command and no accountability - just an algorithm deciding who dies.

That's not defense but the most powerful offense.
It's a dangerous concentration of power.


Trump called Dario selfish and arrogant. 
But Dario Amodei is the only major AI CEO publicly saying this out loud. I've read his 15,000-word essay 'The Adolescence of Technology'. Twice and highly recommend it.

Actually, Dario's company (Anthropic) has the most powerful technology in history of humanity. But he wants it to be used for good. 


The whole problem and rhetoric about this is about who actually gets to decide.

Who should decide how AI gets used in war?

AI, the Pentagon or the CEO of a tech company?

We're in the middle of a pricing overhaul at Juma and wanted to share what we're thinking. 

The problem:
We sell an AI agent for marketing tasks. Until now we've been seat-based ($25-40/seat/month). But as the agent gets more capable, the business model breaks in two ways:

1. Our costs are usage-based, our revenue isn't. 
Some customers pay $40/month and burn $800 in compute. Others pay $250/month and barely use it. The misalignment is getting worse as agents do more. We are not aligned with power users but with sleepers, which is ridiculous as a concept.

2. Seats shrink, not grow. 
One person with a good agent does what 3-4 people did before. Our customers are literally reducing seats because fewer people need to touch the work. We're selling units of something that's deflating (seats).

What we are thinking:
Low workspace fee + unlimited users + credits. 

Heavily inspired by Lovable's model (they hit $200M ARR in 13 months after removing per-seat pricing). 

The idea:
- Free tier with enough credits to complete a few full agent runs (the 'aha' moment)
- One paid plan - flat workspace fee, unlimited seats, credits included, top-ups when you need more
- Enterprise for SSO/compliance/volume

Image shows what we are thinking (far from final).

We are showing what credits can do, because it is the main objections right now. 

We are excited about virality of this because if adding a teammate costs $0, every person invited is a potential power user who drives credit consumption.

What we're still debating:
1. Base price. $29? $49? $99? 
Lower = more viral, higher = better signal + filters out tire-kickers. We don't have a sales team so PLG has to do the work.

2. Credit anxiety..
Everyone on the team has experienced it — you're mid-workflow and suddenly worried about burning credits. UX has to solve this (never break a run mid-execution, show cost before action, auto-recharge).

3. Migration
What do we do with existing clients?

Curious about any thoughts you have whatsoever. 
We need to decide ASAP and get it live by 23 March.

Yavor Belakov was about to give an awesome lecture on agents but something came up. Whoever is coming will get the 2nd best option- me.

If you enjoy my lecture on agents, I'll stop accepting the fundraising talk invitations and will pivot to deep tech lectures.

17 March during AI in Practice Masterclass by Kapital.

↗️ 𝗬𝗼𝘂𝗿 𝗻𝗲𝘅𝘁 𝗔𝗜 𝗶𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁 𝗰𝗼𝘂𝗹𝗱 𝗯𝗲 𝗮 𝗴𝗲𝗻𝘂𝗶𝗻𝗲 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲 𝗼𝗿 𝗮𝗻 𝗲𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 𝗰𝗵𝗮𝘁𝗯𝗼𝘁 𝘄𝗶𝘁𝗵 𝗮 𝗯𝗲𝘁𝘁𝗲𝗿 𝗹𝗮𝗯𝗲𝗹.

Everyone calls their product an "AI agent" now - from simple automation tools to systems that can actually think, plan, and deliver results on their own. When everything sounds the same, how do you tell what's real from what's just marketing?

Iliya Valchanov teaches 1.5M+ students on the topics of AI, data science, and machine learning. He is co-founder and CEO of  Juma (ex Team-GPT), a marketing superagent that executes complete workflows autonomously - analyzing data, generating content, and delivering ready-to-use assets.

At 𝗔𝗜 𝗶𝗻 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗲 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗹𝗮𝘀𝘀 on March 17, Iliya will explain:
➤ How the most capable AI agents actually work
➤ What makes an AI agent "real" today and how to spot the difference between a genuine agent and dressed-up automation
➤ How the best agents plan, do the work, and check their own results 

A 𝗰𝗹𝗲𝗮𝗿 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 for understanding what's under the hood  so you can make confident decisions whether you're building or buying.

Iliya’s keynote is for leaders who want to invest wisely in AI agents without getting lost in the hype.

📅 March 17 | Grand Hotel Millennium Sofia
🎫 Secure your seat → capital.bg/en/event/336


#AIinPracticeMasterclass #AIinPractice #AIAgents #MarketingAI #AgenticAI

Next week Juma will have a low-key launch of something we've worked on for the past 3 months. 

You'll hear a lot more about it soon but this will be special because it's an AI Engineer Foundation meeting in our offices!

I love that this coincides with a long-time member giving an exciting lecture.

Georgi Alexandrov built an autonomous agent that monitors his home server - crash detection, log analysis, auto-recovery. He'll walk through how agentic frameworks work: the tool loop, system prompts, lifecycle hooks, and safety gates. Built with Pi-mono (the engine behind OpenClaw) and a self-hosted 35B model from his wardrobe.

Come and let's talk about agents: 
https://lnkd.in/eV3dtixX

Tech Lead at Sourcelab

What makes a great marketer? Think: Steve Jobs, Elon Musk, Kim Kardashian, most recently even Timothee Chalamet..

I think it's: 
- strategic thinking
- 6th sense for what works
- curiosity and willingness to experiment
- understanding of marketing mix and multi-touch interactions (even on a subconcious level is enough)
- taste!!!

What most great marketers can't do:
- Run ads
- Edit videos
- Set-up Google Analytics
- Write SEO-optimized articles
- Know where to find stuff in HubSpot lol


Keep in mind:
-> The first list will remain human
-> The second list is being automated as we speak

Juma is a superagent that can do any marketing task.

We have rolled out the full capabilities to only 5 workspaces (~100 people). 

If you want early access to try it out, send me a DM.

Or comment 'superagent' below.

Our product can do the same as Lovable, Manus, etc... just 6 months later.

The best companies are best at... speed. 

Speed is the only thing that matters in AI right now.

Memory for agents!

Round led by an OpenAI co-founder, amazing Vasilije Markovic!!!

Founder @ Cognee inc | Semantic memory for AI apps and Agents

We raised $7.5M to scale our knowledge engine that gives AI systems memory.

LLMs are powerful, but they’re stateless — once a session ends, context is gone. Companies have their own context, rules and practices. There are no solutions out there that can bring this knowledge reliably to the LLMs.

At cognee, we’re building a knowledge engine that allows AI systems to retain context, understand relationships, and dynamically update the knowledge over time. 

We raised $7.5M in seed funding led by pebblebed — founded by Pamela Vagata (ex-founding engineer, OpenAI) and Keith Adams (founder, Facebook AI Research Lab) and the amazing Tammie Siew — with participation from 42CAP and Vermilion Cliffs Ventures and angels like Jan Oberhauser from n8n, Charles Dolan, Alexander Dean, Matthaus Krzykowski and Combination VC, and Roman Stanek, Jan Ebbinghaus, Tony S. Kim, Mehdi Ghissassi, Steffen Kirchhoff, Daniel Kirch, Ilya Kozlov, Bojana Miloradovic Parman, Dr. Pascal Wichmann, Jens Lapinski, Bob van Luijt to scale up engineering, accelerate research, and increase number of cognee deployments to clients like Bayer from 70+ to 700+.

I want to thank the team, the open source community, and to our investors for the trust.

If you are struggling with memory, our devs can get cognee running with your data in 7 days. Don’t hesitate to reach out to us.

5 years ago we dropped this ad.

It didn't convert well.

But boy did we enjoy making it!

Може би в топ 3 на клиентите, които са помагали за развитието на Juma → Simeon Penev и VertoDigital

Лекцията ще си струва!

n8n Ambassador | Data & Analytics Consultant at VertoDigital

🤖 AI инструментите вече не са "nice to have".

Миналата година VertoDigital мина от разпръснато ChatGPT използване до 90%+ AI adoption в екипа. Ключов елемент? Juma (ex Team-GPT)

На 26 март от 19:00 ще покажа на живо как използвам juma.ai в ежедневната си работа. Без теория - само практика.

За всеки пример ще видите "преди и сега":
➡️ Колко време отнемаше?
➡️ Какво се промени с AI?
➡️ Какъв е реалният резултат?

Семинарът е безплатен и подходящ за всеки в дигиталния маркетинг.

👇 Линк към семинара в първия коментар

'Am I in the market for better janitorial services?'

I'm still trying to figure out if they want to mop my floors or prep us for an IPO through a leveraged buyout.

taplio