Autonomous AI agents that analyze data and act independently to solve complex tasks have the potential to revolutionize businesses across industries, but in reality, many of today’s AI agent tools either sacrifice functionality for user-friendliness or are too complex for the non-technical users who could benefit most from them.
Michael Nefedov is building a bridge to close that gap, helping businesses with limited technical expertise achieve the true promise of AI.
As co-founder and CTO of next-gen automation platform OmniSales, Nefedov is leading the development of a no-code platform that enables the automation of tasks and workflows across tools like CRMs, email platforms, and spreadsheets. Behind the platform is his breakthrough implementation of remote model Context Protocol (MCP) clients, allowing agents to operate entirely in-browser without depending on a desktop installation or local data sources.
For Nefedov, the mission is clear: collapse the complexity of AI into usable systems that anyone can control.
The Foundations of a Builder, From Reinforcement Learning to OSINT
Michael Nefedov’s technical foundation was forged at the Technical University of Munich (TUM), where he pursued a bachelor’s degree in computer science with a focus on robotics and machine learning.
While still an undergraduate, he convinced his advisor to let him take on a thesis typically reserved for graduate students which involved applying reinforcement learning (RL) to robotic manipulation.
Most systems can only learn one task at a time and struggle when switching between different actions, but Nefedov’s thesis project designed a smarter training method that taught the robot to plan its actions at a high level without neglecting the details of each step, increasing the speed and adaptability of robots learning how to handle new or unique situations.
Nefedov launched his first startup while still at TUM, JukeJar, a “modern-day jukebox” designed for bars, clubs, and private events which lets event guests vote on music playlists in real-time. Just four weeks after being built, JukeJar was part of a collaboration with the Hard Rock Hotel in New York City which saw it process 150+ song requests in a single night.
Later that year, Nefedov began an internship at Synoptic, a startup that analyzes publicly available data like social media posts and news articles to uncover hidden patterns or coordinated activity.
Synoptic enables the crowdsourcing of intelligence on global news, stock markets, and geopolitical events by flagging unusual activity such as sudden spikes in news coverage on a specific topic or coordinated messaging patterns.
Nefedov built machine learning models that could sift through massive volumes of online content to identify suspicious behavior and identify coordinated social media campaigns, finding patterns like multiple accounts posting the same message.
From Startup Lessons to Scalable Infrastructure
Nefedov eventually left Synoptic to shift his focus to voice technology.
He co-founded Fanya, a platform that lets online creators train AI voice bots to speak in their tone and personality. Fanya gained early traction but a sudden disruption in its payment processing system derailed the project.
Around that time, Nefedov met Sezer Kemer, who was studying at HEC Paris. They quickly formed a close bond and a shared vision: what if the power of AI agent tools, tools that can reason, act, and automate across systems, could be made accessible to anyone, not just developers?
The key would be a new kind of architecture, combining the flexibility of large language models (LLMs) with a model context protocol (MCP), which introduces a standard way to connect tools to the agents (LLMs), making connecting a large number of tools easier.
That idea became OmniSales, a no-code automation platform that lets users create intelligent agents capable of performing tasks and workflows across CRMs, spreadsheets, email platforms, and internal databases.
Users describe a goal for OmniSales in natural language, like “follow up with new leads who haven’t responded in three days,” and the platform handles the rest. Behind the scenes, an LLM interprets the user’s intent, while MCP enables the agent to coordinate actions across the tools connected to the platform.
Funding
Within a week of launching, OmniSales received $500k in funding at a $5 million valuation from Frst, one of France’s leading venture capital funds, and both Nefedov and Kemer dropped out of school to focus on the platform full-time.
Nefedov’s core innovation was bringing MCP to the browser. Traditionally, using the framework required local desktop installations and users could only access data from other systems installed on the same desktop device. To solve this, he built one of the first remote MCP implementations that can run entirely in-browser, allowing agents to operate without any setup and connect to other cloud-based systems in real-time.
He also led the design of OmniSales’ backend architecture, which supports persistent agent state (so agents remember what they’ve done), task chaining (so they can complete multi-step processes), and modular execution (so each task can be handled independently).
It’s a complex infrastructure wrapped into an intuitive interface that any business can use, regardless of their technical acumen.
A Mission to Build Developer-Grade, User-Ready Infrastructure
In many ways, OmniSales is the culmination of Nefedov’s entire career.
His dual-policy reinforcement learning system at TUM laid the foundation for modular task execution, his anomaly detection models at Synoptic shaped how OmniSales agents track and respond to unexpected data, and his experiences with JukeJar and Fanya taught him how to build systems that are both scalable and resilient.
But in his eyes, OmniSales is not just a product of experience; it’s a sign of where AI agent tools infrastructure is heading. By bringing remote MCP clients to the browser and wrapping them in a no-code user interface, Nefedov has helped to make the technology more accessible in any type of business environment.
He’s built a platform whose simple interface hides, and therefore makes more accessible, the complexity of its underlying architecture.
For developers watching the rise of AI agents and AI agent tools, Nefedov’s work is a reminder that the future of automation won’t be built solely on the algorithms that power them, but by engineers who can both build these algorithms and make them usable by anyone.
A seasoned technology executive with a proven record of developing and executing innovative strategies to scale high-growth SaaS platforms and enterprise solutions. As a hands-on CTO and systems architect, he combines technical excellence with visionary leadership to drive organizational success.
