The AR industry has hit a wall. Despite billions in R&D investments, today’s most advanced AR devices barely last through a morning meeting. Apple Vision Pro dies after 2 hours, Magic Leap 2 stretches to 3.5 hours, while Microsoft’s HoloLens 2 manages around 3 hours of active use. With IDC projecting the AR market to reach $209 billion by 2025, the path to mass adoption remains blocked by a fundamental paradox: the more data these devices process to create compelling AR experiences, the faster they drain their batteries. As major tech companies race to solve this challenge through exotic battery chemistry and hardware optimization, the real breakthrough might be hiding in an unexpected place—the data architecture itself.
Dmitrii Volykhin, Tech Lead for Wearables Power at Meta and the directly responsible person for AR glasses energy optimization, brings a unique perspective to this challenge. Before joining Meta’s secretive Orion project—the company’s first true AR glasses—Volykhin spent years building systems that processed a billion events daily for advertising platforms, achieving a 35% reduction in customer acquisition costs for major e-commerce players. Now, he’s applying these big data principles to extend battery life in Meta’s wearables division. His Regression Detection Platform automatically detects energy drain patterns across thousands of test runs, while Power Monitor Tools have democratized energy profiling for every engineer on the team. In this exclusive interview with Dataconomy, Volykhin reveals how lessons from high-volume ad tech are reshaping AR’s energy architecture, why treating computational operations like financial transactions could unlock all-day battery life, and what his work on Meta’s next-generation AR devices means for the industry’s future.
From ad tech to AR: Why a billion events matter
The connection between advertising technology and AR energy optimization might not be obvious, but the underlying mathematics are strikingly similar. When Volykhin joined OHM Agency as the first and lead engineer to build SmallData—a mobile marketing platform processing more than a billion data points daily—he was already working on optimization problems that would later prove crucial for augmented reality.
SmallData, designed to manage advertising campaigns across Google, Facebook, and Twitter, faced a central challenge: processing vast volumes of real-time data while staying cost-efficient. For e-commerce leaders like Ozon, often called “the Amazon of Russia,” every millisecond of processing translates into infrastructure costs. For Delivery Club, the country’s top food delivery service, inefficient data handling meant higher customer acquisition costs that threatened unit economics.
The results were measurable. Dmitrii’s architecture cut acquisition costs by 35% for Delivery Club and supported a 314% increase in order volumes.
Google highlighted these achievements on ThinkWithGoogle, recognizing the platform’s innovative approach to real-time ETL. But beyond the numbers was a philosophy.
“When you process a billion events daily for Ozon or Delivery Club, you learn to distinguish which data is critical right now and what can wait,” Volykhin says. “In AR glasses, that same principle saves up to 30% of energy—not all sensors need to work at maximum frequency constantly.”
This insight—that not all data is equal—becomes even more critical for wearables. In advertising, wasted processing costs money; in AR glasses, it costs milliwatts. The constraint shifts, but the optimization problem remains the same: maximize value, minimize resources.
Technical parallels run deeper than prioritization. In advertising, batching reduces database overhead and API calls. In AR, batching sensor readings reduces processor wake-ups—one of the most energy-expensive operations in mobile computing. The real-time synchronization engine Volykhin built to coordinate multiple ad platforms now informs how AR subsystems communicate without draining the battery. SmallData processed a billion events across distributed servers with virtually unlimited power; modern AR glasses must handle comparable volumes on a battery smaller than a car key. The principle—efficient data architecture—remains constant.
The tools that changed everything: Inside energy revolution
When Volykhin joined Meta’s Wearables organization, the company was already deep into developing its AR ecosystem. But the Power and Performance team faced a critical challenge: how do you identify energy inefficiencies across multiple device types—Ray-Ban Meta smart glasses, wrist devices, and the highly confidential Orion project—when each generates thousands of test reports daily?
The regression detection platform
The answer came in the form of an automated intelligence system that changed how Meta approaches energy optimization. Volykhin’s Regression Detection Platform automatically extracts battery insights directly from bug reports during end-to-end performance testing in the laboratory.
What makes this approach effective is its scale and automation. Traditional battery testing involves engineers manually reviewing logs, comparing test runs, and hypothesizing about anomalies. With thousands of test scenarios across multiple devices, this process could take weeks to identify a single regression. The platform transforms this into a data mining problem, automatically collecting and analyzing patterns across the entire testing pipeline.
“Before the platform, engineers could spend weeks looking for the cause of battery drain in a specific scenario,” Volykhin explains. “Now the system automatically analyzes patterns from thousands of test runs and identifies probable causes in minutes. It’s like switching from manually searching for a needle in a haystack to using a metal detector.”
The platform has become what Volykhin describes as “a cornerstone of our end-to-end testing process.” By automatically collecting data on regressions in battery life, it enables the team to address issues long before they reach production.
“Think of it like having a continuously running health monitor for your device’s energy consumption,” Volykhin adds. “Instead of waiting for users to complain about battery life, we catch energy regressions the moment they appear in our codebase. If a new feature suddenly causes a 10% increase in power draw during specific interactions, we know about it within hours, not weeks.”
Democratizing power measurements
The second transformation came from recognizing a fundamental bottleneck: energy measurement was a specialized skill requiring expensive equipment and hardware expertise. Volykhin spent two years developing the Power Monitor Tools to eliminate this barrier entirely.
Before this innovation, measuring power consumption required specialized equipment costing upwards of $10,000, plus the expertise to operate it correctly. Engineers would need to schedule time with the hardware team, set up complex testing rigs, and interpret oscilloscope readings. This created a natural bottleneck—only a handful of specialists could perform these measurements, meaning most code shipped without any energy profiling at all.
The Power Monitor Tools changed this dynamic completely. Now, any developer can measure the power consumption of their code with a single terminal command. When every engineer can see the energy impact of their changes in real-time, optimization becomes part of the natural development flow rather than an afterthought. The tools empower developers to independently conduct experiments and analyze device power consumption, transforming energy optimization from a specialist discipline into a shared responsibility.
For Meta’s wearables division, scaling the Regression Detection Platform across multiple devices while allowing every engineer to measure power consumption has created a cultural shift where battery life is no longer just a hardware problem—it’s a software quality metric as fundamental as latency or memory usage.
The data-first future of wearables
The AR industry faces a fundamental choice: wait for a battery revolution that may never come, or revolutionize how we think about the batteries we already have. Volykhin’s work at Meta suggests the path forward lies in the latter.
Consider the math that defines this reality. Battery technology improves at roughly 5-8% annually—a crawl compared to Moore’s Law doubling processor capabilities every two years. By 2030, processors will be 8 times more powerful, while batteries will have improved by maybe 50%. The traditional response has been to throttle performance, disable features, or accept impractical battery life. Volykhin’s approach suggests a fourth option: radical efficiency through intelligent data management.
“When your glasses die in the middle of the day and you realize it’s because of unoptimized code you could have fixed—that’s the best motivation,” Volykhin says, speaking about his experience wearing smart glasses with prescription lenses daily. “Especially when you know that for people with vision problems, these AI features aren’t just convenience, they’re a necessity.”
What’s emerging at Meta is a new development paradigm where energy becomes a first-class citizen in system design. Software architects debate whether a feature is worth its milliwatt cost. Product managers weigh user value against battery budget. QA teams track energy regressions with the same rigor as crash rates. Features that seemed impossible under traditional power constraints become viable through intelligent scheduling. AR glasses can run complex AI models—but only when the user actually needs them. Background tasks cluster into efficient batches. Sensors dynamically adjust their sampling rates based on context.
“The industry has been waiting decades for a breakthrough in battery technology,” Volykhin observes. “But if you look at the history of tech, breakthroughs more often come from smart use of what we already have. We can’t change the physics of lithium-ion batteries, but we can radically change how we spend every milliwatt.”
The most significant change might be cultural. A generation of engineers trained on Volykhin’s tools will enter the workforce thinking about energy from day one. They’ll instinctively batch operations, question sensor polling rates, and profile power consumption before their code reviews. What Meta pioneered for AR could become standard practice across all of mobile computing.
As Project Orion moves toward launch and competitors scramble to match Meta’s efficiency gains, the lesson from Volykhin’s journey becomes clear. The companies that dominate the next decade of wearables won’t be those with the best batteries or the most efficient chips. There’ll be those who learned that when processing power doubles every two years, but battery capacity barely budges; the only sustainable strategy is to apply the same rigorous optimization that once saved millions in advertising costs to saving milliwatts in AR devices. The billion-event problem taught the industry how to handle scale. Now it’s teaching us how to handle scarcity.
