1. Tell us about your background and what attracted you to SRI.

After earning my Physics Ph.D. from Penn State, I had the privilege of interning at SRI (then Sarnoff Corp). A year later, I joined the company as a research scientist and have been here ever since. SRI has been my sole employer and it's easy to see why. The research environment and topics are unparalleled. I collaborate with brilliant minds daily, tackling new challenges and expanding my knowledge. The feeling of accomplishment that comes from seeing our innovations make a difference is rewarding. It's why I've dedicated my career to SRI.

2. SRI has a nearly 80-year heritage. Along the way, SRI has acquired Sarnoff (formerly RCA Labs) and the Palo Alto Research Center (PARC). What can you tell us about SRI’s mission and vision today?

SRI is a nonprofit research institute with a rich history dating back to 1946. Today, SRI's mission is to create and deliver world-changing solutions for a safer, healthier and more sustainable future. We innovate in AI, health, climate, quantum sensing, work and education, security, space and more.

3. Tell us about SRI's metrics: locations, employees, patents or other metrics to give our readers a better understanding.

  • 1,500 employees
  • 50+ start-ups
  • $4B in federal funding in the last decade
  • 500+ projects per year
  • 15 locations
  • $100B+ market value created
  • 63-acre campus in Menlo Park
  • Other campuses in Princeton, Washington DC, Palo Alto, Boulder and Japan
  • 13,000 patents filed
  • 100 active licenses
  • $100M annual funding
  • Nearly 80-year history.

4. Your group studies AI adaptation at the edge — can you describe what that entails?

AI is everywhere now, but AI is still challenging. We are still only scratching its surface. Most advanced AI algorithms need to run on large servers and my work is to make them available on small devices. I’m not only working with software but also hardware and codesign. For example, to run language models on a small IoT device, you need to design hardware that is high-compute efficient and extremely low power (1000x improvement or more). Regarding algorithms, we are working on novel ideas for efficient training and inferencing using limited resources and power, real-time adaptation to the changing environments, continual learning when a task changes and so on. We are lucky that SRI understands these trends and supports my team. We have successfully secured millions of dollars of funding from DARPA, IARPA and others. Our work has also been funded by venture programs and commercial customers.

5. You have worked with many government agencies like DARPA, IARPA and ARPA-E. What are those agencies looking for and what are the common threads tying all these developments together?

I will outline three programs: DARPA IP2, IARPA MicroE4AI and ARPA-E MicroCAM. The common feature that the agencies our team works with is intelligence at the edge, taking performance, device power and size constraints into consideration.

  • DARPA IP2 is a new design addressing in-pixel computing, which pushes computation in the sensor. We leveraged a biological mechanism called saccades to shift the fixation of attention quickly by only processing salient objects. This way, data bandwidth is reduced, power and latency are reduced, but the performance remains unchanged.
  • IARPA MicroE4AI is a design and implementation of an edge computing SDK to support efficient on-device learning and adaptation to changing environments on a single device. We developed federated learning that supports knowledge-sharing among multiple types of devices that run different tasks and process various data.
  • ARPA-E MicroCAM is a design and development project for a household occupancy detector that can last for a few years without changing the battery. The power consumption needs to be extremely small and support multiple sensors for motion, infrared, acoustic and more. Efficient AI engines that can fuse and detect sensor data, AI-based planning and collaboration are also important.

6. Besides your work for the DOD, you’ve been involved with spin-offs and start-ups. It seems like SRI makes a concerted effort to be engaged in R&D and commercialization. Can you describe how important that is at SRI and how it works?

SRI brings technology into the world by fostering innovation and by providing entrepreneurs with the resources, expertise and IP to bring ideas to life and create real-world impact. We have a track record of successful ventures in the areas of space, automation, AI, drug discovery, robotics, geospatial, 3D and virtual environments and more. SRI’s portfolio includes companies that have been acquired by industry giants such as Apple, Bayer, Ford, Salesforce and others.

7. What timeframe are you and your team looking at for your developments? What has you most excited about the future?

Although we have a long-term 10-year vision, SRI is focused on what we can commercialize in the shorter term (three to five years). I lead NeuroEdge, an SRI-funded program that started our edge computing research. We develop automated edge devices with little human involvement — an AI co-pilot to provide suggestions to the edge devices and make quick decisions to changing environments and unknown tasks. We also focus on developing an abstract layer between the edge and cloud servers that know what and when to query. We demonstrate neuromorphic computing-based visual grounding techniques that leverage vector databases and physical space understanding to generate results. These utilize large language models integrated with the physical environment to help systems at the edge make quick decisions with little human involvement.

8. What areas is the DOD investigating with SRI and what are they publicly stating about how these developments will guide future strategies?

Currently, these and other government agencies are interested in the following areas that are related to my research:

  • Photonic-electronic circuit design targeting low power and efficient communication for space applications
  • Real-time RF circuit design that can dynamically reschedule resources and tasks to improve throughput gain and device utilization
  • Novel ideas regarding AI adaptation at the edge and green technologies
  • New solutions for scalable analog network designs that can reach > 1000x compute efficiency gain compared with digital counterparts

9. What is the culture within SRI? What do you expect from the people on your team and how does SRI support your goals and visions?

The best asset here at SRI is our people, who are very intelligent and relentlessly innovative. We tend to share similar personality features. We never stop learning new technologies. SRI’s leadership provides opportunities for people to grow. We also collaborate among different divisions and different labs. Most government projects require collaborations from inside and outside of SRI and we work with universities, commercial organizations and others.

Great ideas are the seeds of success, but it takes strong collaborations to nurture them into flourishing achievements.

10. What else should our readers know about SRI?

With a nearly 80-year legacy, our technologies, research and ideas have a profound impact on every one of our lives — at home, at work, at school and at play. Our diverse teams collaborate across fields and disciplines and take an integrated approach to supporting our government and commercial clients and addressing some of the world’s greatest challenges.