Areas of interest

A career in technology is a bit like rock climbing: you want to have at least a couple of solid grips that you can lean on, while reaching for new areas. My most solid grips have been in management (especially of innovation), data analytics, and in biomedical informatics, but I have worked on practical use cases, technologies, and applications across many fields and industries, as well as on the finance that drives them. 

Today, I am an independent advisor, pretty much operating quietly in the background (I'm well past the game of engaging in activities just so I can list or humble-brag about them). In the Silicon Valley ecosystem, I am especially interested in entrepreneurial ideas and in startups and growth companies, both in horizontal areas as well as in specific industry applications. Part of what energizes me most in this ecosystem is learning about and understanding the financial, market, and growth opportunities across a wide range of technologies and application areas, without becoming a one-trick pony in any one domain or technology. I have worked closely with founders in areas ranging from fitness to fintech, and from sustainable materials to anything with data. Most of my time today I spend on analyzing and assessing companies, looking at their financials and trajectory, market and opportunities, and technical approach and differentiation.

On the healthcare/academia front, I briefly returned to Stanford full-time in 2019 as the Executive Director of the Biomedical Informatics Program, where I did my PhD research myself. I treasured this opportunity to work closely with the students, faculty, and staff in the program, particularly given how much the field has evolved and become central to biomedical research, and to be a member of the staff of the Department of Biomedical Data Science in the School of Medicine, albeit for only a year.

Earlier Research

My earlier enterprise operational/executive and R&D career spans machine learning, data management and analytics, security analytics, IOT, data de-duplication, healthcare (clinical informatics, analytics, electronic health records, and innovations towards enhancing patient safety and the patient experience), personalization, and operational efficiency. 

As a researcher at Stanford University, I conducted research into high-performance algorithms for probabilistic reasoning (funded by the National Science Foundation and the US Army Research Office). We investigated, designed, and implemented pragmatic inference methods for Bayesian networks, investigated how these performed under resource constraints, and developed tools for comparing algorithms empirically using randomly generated as well as carefully engineered probabilistic networks. We also built prototypes in multiple clinical domains including lymph-node pathology, anesthesiology and pediatric oncology. And we toyed around with how to commercialize the results of this work.

I developed a methodology to generate explanations of probabilistic inference results in Bayesian networks based on network topology analysis and sensitivity analysis. I also collaborated with pioneering researchers in anesthesia, developing a tool to study fatigue and distraction among anesthesiologists.

When I started my industrial research career, I was a researcher on a team that developed a Physician’s Workstation and an early electronic patient record prototype (it was targeted to be much more than a prototype, but not surprisingly real-world deployment at scale and comprehensive scope was much harder than we anticipated). I worked on methods and tools to automate the determination of relevance among clinical data items, to create "smart displays." This work was part of an emerging wave of systems that provided comprehensive access to clinical information, with integrated decision support, from heterogeneous data sources, and implemented in a way that is consistent with the clinical workflow during an episode of care. We piloted this work at several leading healthcare institutions. In addition, being part of a leading industrial research group in health tech, I had a chance to explore a broad set of early-stage activities ranging from sleep to patient monitoring. 

Management and Executive Research Interests

We later formed a team with core capabilities in data mining, information management (content analysis), automated reasoning and decision support, machine learning, operations research, algorithms, statistics, and economics. We found a bunch of juicy practical problems related to technology services and customer support - for example, in content categorization and incident quantification (and training the models on very noisy data); server configuration analysis for mission-critical support; self-calibrating monitoring and change-point detection on manufacturing line data; using call-center and support cost information to optimize parts selection for on-site calls; and detecting transmission errors when backing up and transferring large amounts of data across long distances. The latter evolved into interesting data deduplication capabilities, eventually contributing to an entire storage product line. In addition, we worked on hugely impactful operational problems such as procurement optimization, structuring channel incentives, marketing optimization, revenue modeling and prediction, and product portfolio optimization. 

When we had a chance to re-invest in healthcare, we created a portfolio of research projects, demonstrators, and client co-innovation collaborations to develop innovative industry-specific assets and business solutions. Examples included a patient safety dashboard to reduce line infections and other risks in the ICU, an operating room schedule optimization solution, smart sensing for fall prevention and non-obtrusive patient monitoring (sleep, vitals, activities of daily living), and tools to help patients bridge communication gaps with the healthcare team.

Over time, my management and executive responsibilities grew, including leading teams in multiple countries and areas of expertise. I led the company's applied innovation investments in analytics, machine learning, and data engineering targeted at HPE's software businesses (security, big data platforms, application development management, IT ops mgt, and information management & governance). I also led the software and analytics research targeted at HPE’s largest research investment, The Machine, with the aim to lead the transition to memory-driven computing. We created a number of open-source projects, including Apache Trafodion (an enterprise-scale SQL-on-Hadoop solution enabling OLTP workloads with an amazing parallel-aware query optimizer that builds on a great legacy), and Sparkle (a set of performance enhancements for Apache Spark for large-memory workloads). Finally, I partnered with the Stanford University School of Medicine to help them create the first ACGME-accredited fellowship program in clinical informatics.

Entrepreneurial and Investor Life

While I briefly dabbled with startup life as a grad student, I then joined HP Labs right after school. It led to a great career at HP, but also to the weird situation where I lived and worked in the middle of Silicon Valley, saw and participated in lots of early-stage innovation and great technical breakthroughs, learned the ins and outs of business, finance, and management, and worked my ass off, but always did so from a large enterprise environment. So when the opportunity emerged to re-join the wild and crazy real valley ecosystem, it was the chance of a lifetime. 

Since then, I've seen thousands of pitches and worked closely with many founders. I've gone through the agonies of due diligence and the difficult decisions that investors must make based on limited information about a company, their trust in the founders, their beliefs about technical and market trends and validation of claims made. I've developed my own version of a "nose" for companies I like and the ability of this one to execute on their vision, and I have a pretty good BS detector.  I am pretty open-minded about the application domain and technical stack and love learning about new ones, although I have my own heuristics and biases. And I have learned (often the hard way) that good ideas and good technologies are cheap and plentiful, while founders and companies that can execute on them are the real thing.