Alan J. Nawoj
Engineer, tech startup founder, and independent researcher in optimization, embedded systems, and machine learning.
San Diego, CA.
About
I am an engineer and entrepreneur with an interest in applied machine learning, wearable electronic devices, and mathematical optimization as they relate to human health and athletic performance. I spent more than two decades building software and hardware systems, most recently as founder and CEO of Beacon Mobile, an automotive IoT and mobile technology company that I bootstrapped from the start to an ultimate acquisition by Vontier Corporation (NYSE: VNT). My educational background includes an M.S. in Computer Science from Harvard and a B.S. in Electrical and Computer Engineering from Cornell.
I am now returning to academia to deepen my knowledge of the theoretical and applied aspects of embedded systems, optimization, and machine learning. My current interests span several areas within EE and math, but I have a particular interest in how optimization techniques and adaptive machine learning models can be synthesized directly into low-power embedded firmware to predict physiological degradation, optimize metabolic output, and prevent athletic injuries in real time.
Selected Research & Academic Projects
Momentum-Based Variance Reduction in Non-Convex SGD
UCSD ECE 273 · 2026- Analyzed convergence bounds for non-convex stochastic gradient descent, focusing on STORM and its use of momentum for variance reduction.
- Hypothesized that STORM converges slower when large gradients occur early in a run, and tested this by adding an exponential forgetting factor (EFF) to the denominator of its adaptive learning rate.
- Extended the idea to a time-varying EFF. In Python simulations on a standard non-convex benchmark, the variant consistently reached a given epsilon stopping threshold faster than the original algorithm.
Binomial Distribution Computation in R
UCSD Math 189 · 2026- Traced how R computes binomial probabilities accurately across very large sample sizes despite factorial overflow and floating-point limits.
- Derived the probability mass function logic step by step, showing how Stirling's approximation and its error term, computed via an asymptotic series, resolve the numerical problems.
Experience
Founder & CEO / System Architect, Beacon Mobile LLC
2009 – 2025- Solo-founded and bootstrapped an automotive IoT and mobile software platform; scaled the backend to millions of transactions across three continents for large enterprise brands.
- Sole system architect for the full hardware and software stack, personally writing over 600,000 lines of code across embedded firmware, AWS microservices, and iOS, Android, and web clients.
- Designed and deployed hundreds of custom microcontroller-based IoT devices with industrial sensors and I/O, interfacing with disparate legacy equipment.
- Filed US patent application as sole inventor for an edge-computing method that uses connected-car telemetry to manage car wash memberships.
Software Architect / Technical Team Lead, Charles River Development
2009 – 2015- Led two engineering teams and developed high-performance portfolio risk algorithms in Java and C#, built on mathematical frameworks such as Principal Component Analysis (PCA).
Software Architect / Technology Lead, Deloitte Consulting
2004 – 2009- Architected Java and C# web applications for US and South African public sector clients, and established development frameworks reused by later projects.
Associate / Software Engineer, Booz Allen Hamilton
2001 – 2003- Wrote high-performance C code for a data aggregation and analysis application used by multiple US defense intelligence agencies. Held Top Secret Security Clearance.
Education
Harvard University
2004Cornell University
2001Continuing Education & Certifications
University of California, San Diego
2026Coursera / DeepLearning.AI / Stanford Online
2025Publications & Patents
- Nawoj, A. J. Method for Managing Car Wash Membership Programs Using Unique Vehicle Identifiers from Connected Cars. US Patent Application 16/741,432, filed Jan 2020.
- Nawoj, A. J. Parallel Processing in Java-based Web Applications. Deloitte Consulting Java Technology Group Newsletter, Feb 2009.
- Lorincz, K., Malan, D., Fulford-Jones, T., Nawoj, A., Clavel, A., Shnayder, V., Mainland, G., Moulton, S., Welsh, M. Sensor Networks for Emergency Response: Challenges and Opportunities. IEEE Pervasive Computing 3(4), Oct–Dec 2004.
- Nawoj, A. J. Energy Tradeoffs between Collection-Based Routing Protocols in Wireless Sensor Networks. Harvard University Technical Report TR-03-04, Feb 2004.
- Nawoj, A. J. and Gorniak, M. J. Ontology Development Challenges and Applications Using the DARPA Agent Markup Language (DAML). Air Force Research Laboratory Technical Report, Sept 2003.
Athletics
I am a lifelong endurance athlete with 39 marathons completed thus far, including five Boston Marathons. In 2025, I finished as the top American and second overall male in the World Marathon Challenge (seven marathons on seven continents in seven days). I previously held the Guinness World Record for fastest aggregate marathon times on seven continents from 2014-2018 and was the overall champion in the 2013 Antarctica Marathon.
Contact
alan@nawoj.org · San Diego, California