AI Infrastructure · Robotics · Physical AI
I build and scale technology where architecture decisions become revenue, operating leverage, and capital efficiency: Azure infrastructure programs associated with roughly $450M in annual recurring revenue; Amazon automation deployed into fulfillment workflows influencing billions of dollars in commerce; and founder-led marine robotics aimed at changing the economics of subsea inspection.
Selected program impact
Experience
My career has stayed close to systems with expensive failure modes and visible economics: cloud infrastructure where compute efficiency shapes margin, fulfillment automation where seconds and downtime become unit cost, and robotics where autonomy must earn its way into production.
Lead complex hardware-software programs across next-generation cloud compute, custom silicon, platform readiness, deployment, capacity, and production execution. The work sits inside Azure's commercial engine, where performance per watt, server utilization, deployment velocity, and platform adoption have direct revenue and margin consequences.
Built and scaled robotics, sensing, and warehouse-automation programs across Amazon Robotics and the founding technology organization behind Amazon Go. Programs included pack and fulfillment automation, production robotics, functional safety, and manipulation systems scaled from pilots into multi-site operations—where throughput, downtime, labor leverage, and cubic utilization were financial metrics as much as engineering metrics.
Co-founded Mefferdi to build autonomous marine systems for inspection, research, and persistent operations. The commercial thesis is straightforward: reduce the dependence on crewed vessels, specialist operators, and repeated manual missions by pushing perception, control, and decision-making to the edge.
Developed an early operating foundation in vehicle electronics, embedded systems, component tradeoffs, manufacturing constraints, and safety-critical engineering—experience that later carried into robotics and cloud hardware at much larger scale.
Selected work
I am most interested in the point where technical performance becomes an economic system: price-performance in compute, units per hour in automation, downtime and maintenance in robotics, or mission cost in autonomous marine systems.
Translate platform architecture into price-performance, fleet efficiency, capacity, adoption, and revenue—not simply benchmark wins.
Move robotics from pilots into operating networks where throughput, utilization, downtime, labor, and safety determine whether the technology deserves to scale.
Worked on systems where perception quality was inseparable from customer trust, transaction accuracy, operating cost, and the economics of scaling a physical store format.
Build edge AI, perception, control, telemetry, and robotics around a commercial objective: make inspection and persistent underwater operations cheaper, safer, and easier to deploy.
Public record
Public sources do not disclose every internal program metric. They do, however, establish the commercial and operating scale of the platforms, technologies, and organizations in which this work sits.
Microsoft reported that Azure surpassed $75B in revenue in FY2025, up 34% year over year.
Microsoft FY2025 Annual Report ↗Microsoft reports Cobalt 100 across 32 Azure regions; internal workloads have seen up to 45% better performance while using 35% fewer compute cores.
Microsoft Azure ↗In 2022, Amazon reported $514B in net sales and $84.3B in fulfillment expense—making automation, throughput, and downtime economically consequential at extraordinary scale.
Amazon 2022 Form 10-K ↗Amazon says robots assist with 75% of customer orders and has deployed more than one million robots across its operations network.
Amazon Robotics ↗Public patent records name Stavan Dholakia as an inventor on optimized package-loading technology designed to improve cubic utilization using 3D sensing.
Google Patents ↗Microsoft's Garage Wall of Fame lists Stavan Dholakia among the team behind a robotics initiative for wildfire protection and mitigation.
Microsoft Garage ↗Mefferdi publicly describes autonomous marine platforms spanning AUVs, ROVs, persistent systems, and an onboard autonomy software layer.
Mefferdi ↗Company-scale figures above provide public context and are not presented as personal attribution. Program-level impact figures on this page reflect operating metrics associated with work I directly led or supported.
Research & service
My research interests center on safe foundation-model agents, neuro-symbolic constraints for robotic systems, sim-to-real evaluation, and reliable autonomy. I also contribute to the broader technology community through technical reviewing, conference participation, patents, and technology evaluation, including service as a CES Innovation Awards judge in advanced mobility and robotics.
Operating principle
A technical result becomes a business result only when it survives scale, economics, reliability, and the people who have to operate it.
Contact
For professional conversations, research, advisory work, and speaking.