Top 10 AI Innovators in Programmatic Infrastructure

The physical ceiling nobody is talking about

The programmatic industry has a problem that goes beyond cookieless and MFA. Total bid volume continues to explode, while the cost of processing it has decoupled from media value. Server DRAM prices are projected to surge by 130%, with memory costs peaking at 23% of the total bill of materials. The Ad Tech Tax is no longer just about the intermediaries taking a cut. DSPs are bleeding capital to process synthetic traffic that should never have reached them.

The leaders on this list are the ones moving the industry into what we are calling the great reconstruction: implementing agentic frameworks that move decisioning closer to the core transaction, pruning the bidstream before it hits the expensive memory layers of the platform.

They represent ten distinct disciplines, from exchange architecture to clean room identity to attention-based waste mitigation. What they share is a conviction that the fix is not incremental. The architecture has to change.

The programmatic ecosystem does not need to be rebuilt for an agentic world. The foundation is already in place and proves its worth trillions of times a day. But the foundation alone is not enough. What is needed now is decisioning at the source.

Meet BRAVE’s choice for Ad tech AI innovators of 2026:

The foundation guardian

The mission: Agentifying AdCOM for an 80% reduction in network latency. Ghanbari is driving the transition from manual rule-setting to autonomous, objective-based bidding at exchange infrastructure level.

The tech win: Spearheading Index Exchange’s pivot toward high-efficiency auction mechanics through the “Agentifying AdCOM” roadmap and modernization of exchange infrastructure for sell-side decisioning, primary contributor to the ARTF protocol and adoption.

Proof of influence: Author of Programmatic’s agentic future needs a stronger foundation (March 2026). Key contributor to ARTF protocol and strategy. Serves on the IAB Tech Lab’s Board of Directors Executive Committee.

The orchestrator

The mission: AgenticOS Orchestration: building the operating system for agentic programmatic workflows.

The tech win: Architecting AgenticOS, a standardized container environment where fraud detection, identity, and deal activation agents coexist. This architecture eliminates the latency tax of external HTTP calls, moving transactions into a high-efficiency autonomous environment. Lead for PubMatic’s Agentic Accelerator Program (March 2026).

Proof of influence: Lead architect of PubMatic’s Agentic Accelerator Program. Featured in the PubMatic Expansion series.

The standard bearer

The mission: Predictive bid routing: ensuring the DSP only processes auctions with a verified probability of reaching a real human user.

The tech win: A legendary figure in OpenRTB, now applying ML to drive predictive bid routing and real-time optimization within Amazon’s retail media network. His fingerprints are on nearly every major open standard in the industry. At Amazon, he is focusing on how predictive modeling can eliminate waste before a bid is even placed, protecting infrastructure at the world’s largest retail media network, serving both brand advertisers and Amazon’s own commerce data simultaneously.

Proof of influence: Legendary figure in the OpenRTB Consortium. Leading implementation of predictive bid routing within the Amazon DSP. Neal also serves as a board member of IAB Tech Lab, where he is a contributor to the development of the OpenRTB protocol, the deal_id, OpenRTB Native, ads.txt, ads.cert, sellers.json, and other associated standards.

The source coder

The mission: Supply-side bid optimization: modifying and throttling irrelevant requests pre-bid to deliver pre-qualified, high-signal traffic to DSPs.

The tech win: Leading OpenX’s 50-person Outcomes & Optimization team, Calabrese has built two AI systems that leverage intelligence on the supply side. Results by OpenX uses deep neural networks to score bid requests in real time across 40+ ID-free signals, routing only the impressions most likely to hit a buyer’s specific goal, delivering an 11x improvement in cost per acquisition. OpenXBuildTM goes a step further, enabling buyers to run their own custom decisioning models inside the exchange before the auction runs. By moving decisioning upstream, OpenXBuild enables real-time, custom decisioning for advertiser-controlled AI with no added latency.

Proof of influence: Featured in the Futureweek “My AI” series on elevating exchange performance.

The hyperscaler

The mission: Compound AI at scale: building the open-web ML infrastructure that closes the performance gap between independent DSPs and walled gardens, one impression at a time.

The tech win: Ahn left Google in 2013 after building YouTube’s video profitability prediction system and serving as founding Tech Lead Manager of the Android Data team. He recognized that the compound AI powering YouTube’s monetization was entirely unavailable to the rest of the mobile advertising ecosystem, and co-founded Moloco to change that. Twelve years later, Moloco’s compound AI system processes more than 5 million bid requests per second across 2 billion consumers and 2 million mobile apps. The architecture is not a single model. It is a layered system of specialized neural networks executing in concert: auction win probability, user engagement prediction, post-install event optimization, and real-time LTV estimation, all running within RTB latency constraints. In April 2026, Ahn extended the compound AI system to connected television with the launch of Moloco Ads for Performance CTV, applying impression-level ML optimization to the living room screen for the first time at Moloco’s scale. Profitable for five consecutive years, Moloco stands as the clearest proof that independent ML-native DSP infrastructure can compete with the walled gardens’ scale advantages without burning infrastructure capital on traffic that should never have reached the bidder.

Proof of influence: Built YouTube’s video profitability prediction system at Google. Ranked 4th globally in mobile app media power by AppsFlyer Performance Index (December 2025). Named Most Exceptional Entrepreneur by Goldman Sachs. Launched Moloco Ads for Performance CTV (April 2026).

The commerce conductor

The mission: Building the commerce intelligence and infrastructure layer for the agentic, outcomes-oriented era of media buying, powered by $1 trillion in annual transaction signals.

The tech win: Parsons joined Criteo in 2020 and has spent five years building what is now the most commercially grounded AI infrastructure play on this list. Criteo’s AI-powered engine has access to more than $1 trillion in annual commerce sales, giving it a view of shopping intent that no other platform can match. In February 2026, Parsons and the Criteo team launched the Agentic Commerce Recommendation Service: a system that connects Criteo’s commerce intelligence to AI shopping assistants, demonstrating 60% better recommendation relevancy than approaches based solely on product descriptions. In March 2026, his team expanded Criteo GO to full self-service and introduced the Onboarding Agent that autonomously configures campaign parameters and forecasts results. That same month, Criteo was named OpenAI’s first ad tech partner for its ChatGPT ads pilot, reinforcing the company’s position at the forefront of AI-driven commerce. His thesis on what comes next is clear: As discovery turns into performance and media buying grows more outcomes-oriented, the next era of media buying will belong to platforms that can shape decisions earlier in the journey and turn that influence into measurable outcomes.

Proof of influence: Leader of Criteo’s unified product and performance media vision for the next era of AI. Featured in Beet.TV’s Commerce, Go, Agent-Ready, and Adexchanger. Launched Agentic Commerce Recommendation Service (February 2026). Named first ad tech partner integrated into OpenAI’s ad pilot (March 2026). Expanded Criteo GO with full self-service access and Onboarding Agent (March 2026).

The gatekeeper

The mission: Real-time bidstream suitability: integrating computer vision and NLP into the pre-bid stream to block unsuitable and synthetic traffic with a specific focus on streaming TV, the fastest-growing and least-verified environment in programmatic.

The tech win: Morello arrived at DoubleVerify in 2022 with an unusual background: AI operations and fraud prevention at TikTok, Amazon, and GoFundMe, not traditional ad verification. He brought that fraud-first instinct to the pre-bid layer, integrating computer vision and NLP into cryptographically verified containers to stop unsuitable traffic at the source rather than measuring it after the fact. His most recent major work is DV Authentic Streaming TV, launched at CES 2026: a natural-language AI planning agent that lets buyers describe brand goals conversationally and enforces them pre-bid across streaming environments. The problem it solves has a number: over $1 billion in misplaced CTV ad spend per quarter.

Proof of influence: Key architect of DoubleVerify’s AI product operations and pre-bid suitability workflows. Led the AI product operations behind DV Authentic Streaming TV, launched at CES 2026.

The attention scientist

The mission: Attention-based waste mitigation: separating human attention from bot-driven waste at scale, using conversational AI and explainable data science to make media quality actionable for every marketer, not just data teams.

The tech win: Leading the data science efforts to separate human attention from bot-driven waste at scale. Her work with the IAS Agent enables marketers to use conversational AI to prune low-quality inventory before it triggers expensive bid requests, protecting client compute budgets. IAS Agent, launched in December 2025 and demonstrated at CES 2026, is an AI-powered campaign assistant built on explainable AI principles that surfaces performance insights up to five times faster than manual analysis — the deliberate opposite of a black-box system. Beyond the IAS Agent, Furze is building the ethical AI governance frameworks that define how media quality intelligence should be developed and deployed responsibly at a global scale. She also recently led recertification for IAS’s ISO 42001:2023 AI Management System.

Proof of influence: Recertification lead for IAS ISO 42001:2023 AI Management System. Webby Finalist for IAS Agent generative AI innovation. Former Global VP of Customer Science at LinkedIn Marketing Solutions. Low-Quality GenAI Avoidance pre-bid tool launched in April 2026.

The Protocol Architect

The mission: Open protocols for the agentic era. Bosko is one of the top voices leading the standards by which AI agents transact across the advertising stack, outside the walled gardens.

The tech win: Milekic is a serial engineer-founder whose 20-year arc through FreeBSD kernel development, AdGear (acquired by Samsung Electronics in 2016), and now Optable has put him at the centre of every infrastructure shift programmatic has undergone. His most consequential current work is the Ad Context Protocol, launched October 15, 2025, which he helped found and now helps steer: a standard for how buy-side and sell-side agentic systems exchange structured data about audiences, inventory, and campaign briefs outside the traditional bidstream. Built to work with Anthropic’s Model Context Protocol, AdCP is positioning itself as the interconnectivity standard for the AI era of advertising. In January 2026, Optable launched the Audience Agent built with AdCP support, automating publisher ad planning by reading natural language RFPs and proposing custom audiences. In March 2026, the Audience Agent was integrated into PubMatic’s AgenticOS, one of the first production-level deployments of AdCP across live programmatic infrastructure.

Proof of influence: Member of IAB Tech Lab Board of Directors (May 2024). Founding member of the Ad Context Protocol (October 2025). Co-author of the OPJA (Open Private Join and Activation) specification. Co-leads the PMPs and Deals working group at AgenticAdvertising.org. Launched Optable Audience Agent (January 2026). AdExchanger named Optable’s ID Switchboard the Best Identity Technology of 2025. Optable named a Major Player in IDC MarketScape for Data Clean Room Technology (May 2025).

The margin defender

The mission: Buy-side path enforcement: defining and piloting Publicis Connected Media France’s agentic AI strategy, ensuring clients buy only from verified, low-latency supply paths that protect margin and deliver durable value.

The tech win: As the first CAIO appointed at Publicis Connected Media, Pfeiffer is implementing agentic optimization to ensure clients only buy from highly verified, low-latency supply paths. Her mandate is to define the agentic and generative AI strategy for one of the world’s largest media buying operations, governing how AI is deployed across buying, influence, and communication strategies. Her appointment arrives as Publicis Group executes its most consequential AI bet: a full-stack agentic marketing partnership with Microsoft announced in April 2026, combining Azure cloud infrastructure, Microsoft Copilot agents, and Epsilon’s identity data layer. Pfeiffer’s role is to ensure that AI serves performance, creativity, and media relevance, not simply automate what already exists.

Proof of influence: Appointed as the first Chief AI Officer for Publicis Connected Media to pilot agentic AI strategies. Named Femme du numérique de l’année at the World Cyber ​​& Digital Women’s Trophies (December 2025).

The technical framework: why ARTF 1.0 matters

For infrastructure leads, the Agentic Real-Time Framework (ARTF 1.0) is the new USB-C of ad tech. By moving logic into cryptographically verified containers, the industry is achieving two structural improvements that were previously impossible at scale.

Sub-100ms communication eliminates the external hop that historically crippled bid performance. Every millisecond saved at the infrastructure layer reduces the compute tax on every auction in the system. At the volume programmatic operates at, this compounds rapidly.

Standardized intents, using specific commands like ACTIVATE_DEALS and BID_SHADE to propose OpenRTB patch mutations to the bidstream, ensure the system only processes high-value traffic. Rather than rejecting bad inventory after the fact, ARTF 1.0 enables the system to define what good inventory looks like before the auction begins.

The shift ARTF 1.0 enables is a structural change. Moving decisioning upstream, closer to the source, is the only way to solve programmatic's infrastructure problem. Share on X

What this means for the market

The executives on this list are not optimizing the existing programmatic stack. They are rebuilding the decisioning layer from the ground up, moving intelligence upstream to the point where it can actually prevent waste rather than measure it after the fact.

The common thread across all ten is a shift from reactive to proactive infrastructure. Fraud is not detected and removed. It is prevented from entering the auction. Identity is not resolved after the bid. It is verified before the request is made. Suitability is not checked post-bid. It is enforced at the source.

For buyers, this means the supply path is no longer a cost center to be minimized. It is a strategic asset to be optimized. The DSPs and agencies that build relationships with the infrastructure layer, rather than treating it as a commodity, will operate with a structural cost and quality advantage that compounds over time.