Speed as Strategy: The AI Infrastructure Behind Smarter Ad Targeting

Photo: Shutterstock

Milliseconds decide who wins in today's digital advertising ecosystem. In such a high-stakes environment, every interaction must deliver intelligent personalization - tailored ads, relevant content, and seamless transitions. In my experience leading infrastructure at a major US tech firm, I've seen firsthand how the industry's vision often collides with the reality of fragmented systems and inherited complexity.

Most advertising infrastructure wasn't designed for the real-time era. Behavioral signals, contextual triggers, and privacy regulations now compete for attention in sub-second intervals. I've managed transformations where legacy workflows and patchwork compliance added more latency than value.

I'm Paras Choudhary, a senior technical program leader in AdTech. I oversee AI-powered infrastructure at a major US tech firm's ad platform - systems that process hundreds of terabytes daily and make contextual decisions near-instantly. I've led cross-functional teams across mobile, streaming, and retail, where adaptability, precision, and trust aren't afterthoughts - they're requirements.

What I learned? We don't need more complexity - we need unified foundations. When velocity and ethics are part of the design, systems stop reacting and start predicting.

This is how we reimagined our infrastructure - and why sub-second decision-making is now one of the most valuable currencies in modern advertising.

The Challenge: Scaling for Hundreds of Terabytes Daily

To deliver personalized experiences in real time, we process hundreds of terabytes of daily behavioral, contextual, and environmental data. These signals - from mobile interactions and CTV streams to web clickstreams - arrive in inconsistent formats and with varying degrees of reliability.

At this scale, complexity isn't just technical - it becomes a direct business risk. Missing a sub-second delivery window means the ad is never seen, resulting in lost impressions, reduced engagement, and missed revenue opportunities. Precision must be both immediate and reliable.

Compounding this, privacy mandates require that consent be assessed and honored instantly, not downstream. Our systems can't afford to lag - every decision must respect real-time compliance rules.

The organizational layer adds another dimension of difficulty. We had to coordinate across over 50 engineering teams and over 10 executive orgs - many of which were built independently and optimized for different priorities. Synchronizing these functions required rethinking standards, aligning incentives, and creating a shared understanding of latency and compliance goals across a sprawling ecosystem.

Figure 1: Real-Time Ad Decisioning Lifecycle - How milliseconds define the path from data to ad delivery

McKinsey's insights into commerce media echo our urgency: success hinges on the seamless integration of data, timing, and intelligence to drive relevance at scale. Translating that insight into action became our mission - rebuilding the infrastructure so each signal, model, and response worked in perfect sync.

The Solution: Distributed AI, Engineered for Velocity

Our breakthrough? We stopped waiting on the cloud. By embedding inference models directly within ad-serving systems, we moved decision-making closer to the user. This shift eliminated central bottlenecks, enabling instant evaluations of context, behavior, and consent on the device.

No more calling home. Our models now assess signals privately and immediately, ensuring high-speed decisions with complete transparency. We also introduced micro-batching to preserve fidelity while maintaining responsiveness under load.

Privacy and compliance weren't afterthoughts - they were embedded in every layer. Region-specific policies, audit logs, and consent flows are orchestrated seamlessly, enabling trust without tradeoffs.

Figure 2: Scale and Signal Flow - A snapshot of the data sources powering audience insights in real time

A performance system must balance automation with adaptability. Our modular design allowed us to push updates rapidly, with zero service disruption.

As Elena Knox of BBDO New York puts it,

"AI can be an incredible creative tool, and if we keep getting in its way of fearing it, it'll only date us."

That mindset informed our approach: AI optimizes and helps shape brand-relevant, context-aware campaigns.

Application: Cross-Surface Targeting Without Lag

Photo:Unsplash

Once the new systems were operational, our focus shifted to scale. The real test was ensuring real-time, personalized, brand-safe delivery across all surfaces - from mobile and tablet to smart TVs and beyond.

Each screen - a smartphone in a pocket or a smart TV in a living room - has latency quirks and signal constraints. To navigate these, we developed orchestration layers that normalize inputs and harmonize creative delivery, regardless of screen, bandwidth, or rendering engine.

The goal wasn't just raw speed but harmony - seamless performance across devices, networks, and creative formats. We built orchestration layers to normalize signals and unify delivery across screen types, ensuring consistency regardless of device, bandwidth, or rendering engine. Each impression had to meet targeting accuracy benchmarks and respect platform rules without introducing lag or risk.

Figure 3: Cross-Surface Ad Flow - Example of synchronized ad delivery across devices

This mirrors findings from the GLBIMR journal, which emphasized the importance of adaptive AI architectures across interfaces to sustain performance and ethical clarity - an idea we implemented firsthand. In our case, that theory translated directly into platform-specific enhancements that respected performance limits while preserving strategic coherence.

The Business Outcome: From 60% to 95% Addressability

The results? Transformative.

Under my leadership, we expanded contextual and behavioral targeting from 60% to 95% - a leap, not a tweak. That jump redefined our reach and advertiser effectiveness across every surface.

Advertisers and partners saw tangible business outcomes, similar to those documented in Single Grain's case studies, highlighting measurable ROI lift through programmatic precision and infrastructure agility:

• Match rates improved across mobile, video, and display.
• Ad relevance improved, driving higher engagement, stronger click-throughs, and more meaningful impressions.
• ROI rose as impressions became more meaningful.
• Privacy controls remained intact, building user trust.
• Operational costs stayed stable despite expanded targeting scope.

Figure 4: Addressability Growth Curve - Impact of AI on audience expansion (2023--2024)

A study on the evolution of programmatic advertising published via ResearchGate validates what we experienced firsthand: ethical targeting systems don't dilute efficacy - they enhance it. Responsible infrastructure design now serves as a market advantage, not a constraint.

Similarly, industry researchers writing in the Journal of Digital and Social Media Marketing have shown that strategic investments in latency-free infrastructure consistently drive campaign success and brand uplift. These findings mirror our internal data, confirming that faster, smarter systems don't just perform better - they become a competitive moat in a crowded AdTech landscape.

Speed Isn't Just Technical. It's Transformational.

Speed doesn't just win auctions - it wins trust. In AdTech, being first to bid is table stakes. Real strategy is first understanding, adapting, and delivering relevance in real time. Infrastructure isn't just backend plumbing in this environment - it's a competitive differentiator.

I've watched infrastructure evolve - from backend plumbing to a core growth engine. We moved from reactive systems to predictive ones, capable of adapting in real time. We moved beyond bolted-on compliance to privacy-by-design. And we proved that speed, intelligence, and ethics can all coexist at scale.

But what excites me most is what comes next. We've proven that real-time, privacy-respecting AI systems can deliver scale and performance without tradeoffs. Going forward, the winners in AdTech won't be those with the most data - they'll make the smartest decisions in the smallest time window, without compromising trust.

In AdTech, the real differentiator isn't data volume - it's velocity. Ultimately, it's not who has the most data, but who can make the smartest move in the smallest window, without breaking trust. That's the future I'm building toward.

That's the future I'm building toward. For teams working at the intersection of advertising, AI, and infrastructure, this demands better models or bigger datasets and faster, more integrated systems built with intention from the ground up. In the milliseconds between signal and decision, modern advertising isn't just activated - it's defined.

About the Author:

Paras Choudhary is a senior technical program leader specializing in large-scale advertising infrastructure and AI-driven targeting systems. With over a decade of experience spanning video, retail, and mobile platforms, he has led cross-functional teams to architect real-time, privacy-first solutions that process hundreds of terabytes daily. His work focuses on building intelligent, high-performance systems that power the next generation of personalized digital advertising.

References:

• Bajaj, R. (2023). Best practices of programmatic advertising: A narrative analysis. Optimization, 15(2): 92--102. https://www.glbimr.ac.in/pages/OJRM-15-2/Chap4%20Vol.%2015(2)%20July-December%202023.pdf
• Bhattacharya, J. (January 2024). Best programmatic advertising case studies for exceptional ROAS. Single Grain. https://www.singlegrain.com/blog/programmatic-advertising-case-studies/
• Brodherson, M., Chen, T., Flugstad, J. & George, Q. (2022). Commerce media: The new force transforming advertising. McKinsey & Company. https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/commerce-media-the-new-force-transforming-advertising
• Huh, J., Nelson, M.R. & Russell, C.A. (2023). Programmatic advertising transparency: A conceptual framework and research agenda. Journal of Advertising, 52(3): 305--317. https://doi.org/10.1080/00913367.2023.2227013