Before the commercial web had learned to sell anything, Jeremy Fain wanted to work on it. He grew up in Maryland, close enough to the federal research world that the internet arrived first as infrastructure rather than spectacle. In science magnet programs, he studied computer science. During an internship at the Army Research Laboratory, he encountered the government side of a network that most Americans had not yet been invited to click. Then the World Wide Web brought images and text into the same window, and a direction became visible.
Fain went to Yale for electrical engineering, graduating in 1999, and began his career as an e-business consultant at PricewaterhouseCoopers Consulting. The assignment now sounds almost quaint: help large companies understand email, websites and the possibilities of connecting their systems through the internet. At the time, it was frontier work performed in sensible office clothes. He later helped start a financial-services internet company and developed digital strategies for major brands at Digitas.
The useful detail is not that Fain predicted the internet. Plenty of clever people did. It is that he kept changing seats around the same table. Engineering showed him the system. Consulting showed him the client. Agency work showed him the message. An MBA in marketing and media from Columbia Business School added finance and management. By the time he reached the machinery of digital advertising, he could hear several professional dialects at once.
Five views of the same machine
A career in the plumbing
At the Interactive Advertising Bureau, Fain worked on policy, standards and practices that allowed a young industry to conduct business without improvising every transaction. His work included the first VAST standard for digital video advertising and the industry's Terms and Conditions 3.0. Standards work rarely earns a dramatic soundtrack. It does, however, teach a durable lesson: markets become large when strangers can agree on the boring parts.
CBS Interactive put him on the publisher side. As director of network solutions, he oversaw data-driven products across a collection of branded sites and managed a P&L of more than $30 million. The work included audience targeting, private exchanges and custom audience products. At Rubicon Project, where he eventually led North American accounts, the view widened again. He was responsible for account teams serving buyers and sellers, and for helping more than 400 media companies and 500 demand partners use real-time bidding and new revenue products.
“We saw the standard bidding algorithms all over the place. That didn't jibe with me.”Jeremy Fain, on the irritation that preceded Cognitiv
The objection was plain. Target customers and Walmart customers behave differently, yet standard bidding systems tended to offer both companies the same underlying logic. Even custom algorithms often required teams to guess which variables mattered. Fain wanted a system that could learn those relationships from each advertiser's own data. The idea was less theatrical than the AI slogans that arrived later. It was also considerably harder.
The friends in the machine
In 2015, Fain founded Cognitiv with Marc Hudacsko and Aaron Andalman. The three had known one another while growing up, moving through the same Maryland magnet science and engineering programs. Hudacsko became chief technology officer. Andalman, who trained in neuroscience at MIT and Stanford, became chief science officer. Their reunion makes the origin story sound almost suspiciously tidy: three childhood friends grow up, collect different kinds of technical depth, and return to build a machine that learns.
Fain is funny about the division of talent. He has described himself as a former coder, while noting that Hudacsko and Andalman were much better. The admission clarifies his value. He had enough engineering to understand the architecture, enough operating experience to recognize the market failure and enough commercial fluency to sell a new approach into a skeptical industry. A translator is most useful when he respects both languages.
The first version of the plan was to build custom models from client data and plug them into existing demand-side platforms. The platforms could not handle the decisions Cognitiv wanted to make, impression by impression. So the team built its own. A rejected integration became an infrastructure company.
The clock is the product
Deep learning is computationally hungry. Programmatic advertising is impatient. An available impression may need to be evaluated in tens of milliseconds, while the large models associated with chatbots are allowed to consider their response for seconds. Cognitiv's technical identity lives inside that mismatch. Its systems have to gather signals, calculate value and act before the opportunity vanishes.
This is why Fain pushes back on the idea that AI is synonymous with a chatbot. Advertising happens too quickly. Cognitiv uses smaller, faster models designed for a specific decision environment. Its pitch is that a brand should have a model trained around its own data and desired outcome, rather than inherit a generic recipe.
The product line has expanded with the language of the market. ContextGPT lets a marketer describe a contextual audience in natural language and continually discovers relevant new content. AudienceGPT, launched in March 2026, applies a conversational interface to dynamic audience creation and can refresh segments as behavior changes. The interface is familiar; the old machinery underneath still has to work at bidstream speed.
That distinction has become more important as identifiers and third-party cookies have lost favor. Cognitiv began a European expansion in 2026, positioning contextual intelligence for a market that had spent years thinking about privacy-first advertising. Fain's timing was deliberate. The company had spent a decade building its model and, in his phrase, finally had the “heft and the revenue” to scale it abroad.
“Everyone thinks AI means a chatbot. But advertising happens in milliseconds.”Fain on building for the speed of the market
The other lane
Long before the founder titles, Fain had another demanding system for measuring progress. He began swimming on a summer team in Montgomery County, Maryland, at six. By twelve he trained year-round with the Rockville Montgomery Swim Club. He qualified for nationals at sixteen and swam for four years in college. He began coaching at fifteen and has taught swimmers from four to sixty-four.
In 2007, Fain co-founded the Trident Swim Foundation with Alan Fishman and Kristin Gary. Its Swimmer Scholar program combines structured practice with tutoring, mentoring, test preparation and college counseling. The three principles are almost stubbornly unfashionable: discipline, self-confidence and teamwork. Nobody is promised a shortcut. The program asks students to become better in the classroom and the pool through repeated, visible effort.
It is difficult not to see the rhyme with Fain's business language. He says technology companies live or die by project management: deadlines, mapped processes, a clear route from A to Z. He values hard computer-science fundamentals in hiring. He once summarized Cognitiv's early approach by saying that one very good developer is worth five. The swimmer's clock and the engineer's benchmark share a certain moral severity. The result is there, whether or not you enjoyed the set.
Yet neither swimming nor company building is solitary in his version. Trident treats an individual sport as a team discipline. Cognitiv began among friends whose strengths were usefully uneven. Fain's own job sits in the relay zone: science to product, product to customer, customer outcome back to model. He is the person receiving one kind of momentum and trying not to drop it.
What compounds
By 2026, the public markers had begun accumulating. Cognitiv and Sargento won a 2025 Digiday award for the use of AI in connected television. Fain and Andalman appeared on ADWEEK's AI Power 50. Fain was named an EY Entrepreneur Of The Year New York finalist. ContextGPT adoption rose, AudienceGPT arrived, and the company took its technology into Europe.
Awards are pleasant punctuation. The more instructive achievement is the sequence. Fain spent years in jobs that revealed different failure modes: clients without internet fluency, publishers with monetization problems, an industry without shared standards, exchanges using generic algorithms, marketers carrying fragmented data. Cognitiv was not a jump away from that career. It was the place where the preceding views could finally be combined.
His advice follows the same pattern. Learning to program trained him to separate the fundamental problem from its symptoms, to map the route and to search for an efficient solution. He returned to business school without discarding the engineer's habits. He moved closer to the transaction without losing sight of the user. Then he founded with people who remembered him before any of the titles.
There is something refreshingly pre-fashion about the whole construction. Fain did not arrive at artificial intelligence through a sudden conversion. He followed code into the internet, the internet into marketing, marketing into auctions, and auctions into models that could learn. The vocabulary changed around him. The problem remained recognizable: understand a complicated system well enough to make one better decision, then do it again before the clock runs out.