Michael Nefedov: The Outstanding Engineer Revolutionizing AI Agents for Everyone

By the time most students were wrestling with their first coding assignments, Michael Nefedov was already designing algorithms that taught robots how to move. At the Technical University of  Munich (TUM), he persuaded his professors to let him take on a thesis typically reserved for graduate students: applying Reinforcement Learning to robotic manipulation.

He not only succeeded but also developed a more innovative training method that allowed robots to plan actions at a high level while managing the details of each step, enabling faster learning and adaptability to new or unique situations. For most, this would be a career-defining achievement.  For Michael, it was just the beginning.

Michael’s résumé is already extraordinary: he has interned at Salesforce and Amazon before the age of 21, secured a DevOps job during his first semester, and developed side projects that quickly evolved into startups. Yet, despite standing out in some of the world’s most selective companies, he felt restless.

“I realized I didn’t want to climb the corporate ladder,” he reflects. “Even in these prestigious jobs, I couldn’t see myself there long-term. I needed to build.” Build he did.

His first solo startup, JukeJar, became a “modern-day jukebox” for bars, clubs, and events.  Within weeks of launching, it was deployed at the Hard Rock Hotel in New York City, where more than 150 song requests flowed through his system in a single night. The experience cemented his belief in the power of rapid execution. His next venture, Fanya, gave creators the ability to build AI voice bots in their own tone and personality, generating $18,000 in just three weeks before payment system disruptions stalled its growth. These early attempts taught Michael lessons in scalability, resilience, and adaptability that would become invaluable later.

Michael also contributed to Synoptic, a startup using machine learning to analyze public data from social media and news sources. His anomaly detection models helped uncover Russian propaganda campaigns on Twitter, work that supported Synoptic’s $15 million funding round.  This experience shaped his ability to design systems that can track and respond to unexpected data patterns, a skill that now lies at the heart of his AI agent architecture.

In 2025, Michael joined forces with Sezer Kemer to found OmniSales (OS Recovery). They started out by building technologies that enabled AI agents to communicate more reliably with other applications and tools. They recently decided to apply this tech to one of the least glamorous industries imaginable: debt collection.

Traditional AI assistants are either too simplistic or too complex, making them undeployable in the debt collection vertical. Michael and Sezer went deep to understand the daily problems of collection agencies and how to apply their knowledge of technology to solve them. Their first solution, a voice AI debt collection assistant, revolutionizes the entire portfolio pricing by unlocking accounts with small balances that were previously uncollectible due to the high costs of

human operators. Michael created a scalable and cost-efficient voice assistant that conducts routine calls while being human-like. Another common issue with AI assistants is their compliance and reliability. Michael overcame this challenge by utilizing advanced prompt guards and fine-tuning a custom model for identity verification. After successfully deploying their voice assistant to one of the top 10 US credit unions, Michael and Sezer are now working on portfolio management agents to reimagine how debt collection firms operate.

OmniSales quickly drew investor attention. Within a week of launch, the company secured  $500,000 in funding at a $5 million post-money valuation from one of France’s top venture funds. Both Michael and Sezer left school to dedicate themselves fully to the platform. For

Michael, OS Recovery represents the culmination of years of iteration, from robotics and anomaly detection to consumer apps and AI voice technology.

“Startups are about testing ideas fast, but this one feels different,” he reflects. “It’s not just technology for the sake of it. It’s solving real problems for many businesses in a multi-billion-dollar industry.

Michael Nefedov’s story is about more than technical skill. It’s about daring to leave prestigious paths, embracing the scars of failure, and relentlessly pushing the boundaries of what AI can do.  From robotic learning at TUM to scalable AI at OS Recovery, his journey shows what happens when a talented and outstanding engineer refuses to settle for ordinary solutions.

Have you read?
Countries with the Highest Age Dependency Ratio.
Best Countries to Hide Money.
World’s Most Productive Countries.
Countries by Average Wealth per Person.
Countries With The Most Serial Killers.