Research Portfolio

This profile provides descriptions of several technologies we have developed or worked on. It has a research focus as we come from a research community. We do not expect our customers to present us with such fundamental research problems. But we think it is important to our credibility that we have, in the past, focused on data science problems that were not feasibly addressed with standard tools. We've always been a place where people went when they had problems that were harder than the standard problems or more niche than that. The world has changed. We think it's exciting that data science is suddenly interesting to a lot more people, many of whom don't have the resources of a nation state to work those problems. And the standard tools have gotten a lot better. We look forward to working on these new problems. But we have a good research pedigree as well.


There is a "Printable Version" link at the bottom of each page of the profile to make hard copy a bit easier. Each technology in the portfolio uses a communication tool called a "Heilmeier Catechism" to describe these capabilities. The Heilmeier Catechism provides answers to a very formal set of questions that can be used to describe a technical capability, the problem it addresses, why the given approach is superior to others, etc. The Heilmeier questions have been widely used for this purpose since the 1970s and are still a popular way to describe a solution, particularly R&D solutions to difficult problems. 


We have updated the Heilmeier questions slightly, to add context for Technology Readiness Level (TRL). TRL is well understood in the research community, and provides a scale for determining how much work is needed to make the technology ready for prime time. Higher TRL tends to mean less time to market. Lower technology tends to mean more opportunity for intellectual property. 


Some Common Threads:

If we had to provide a brief summary of our work around some repeated themes, they would be these:
  • Solutions with General Applicability: We seldom work on point solutions. Most of the technology we create has application to multiple areas. There may be uses in defense, but also perhaps in online brokering, or bioinformatics or energy, and as a team we have worked on projects ranging from atmospheric phenomenology to natural gas pipeline metering to image analysis.
  • Getting Reliable Results from Unreliable Pieces: Sometimes the unreliable pieces are equipment and sometimes the unreliable pieces are the information that the equipment processes. The problem is fundamental in statistics, computer science, and engineering. Unreliable equipment is common in settings where the equipment must be located where the information is, rather than where its support system is. To illustrate, a large data center may located where power is reliable and physical security is easy. A military sensor, on the other hand, needs to be located where the militarily relevant information is. That often places it where access is hazardous and the sensor is likely to break or be broken by an adversary. Nonetheless, systems built up around that sensor must continue to operate when the sensor fails. Unreliable information is often harder to deal with than unreliable equipment. How do you reason about conflicting pieces of evidence? How do you build a system that can autonomously perform its job even when competitors subject it to deceptive information? How do you reason about evidence that is obviously related to the goal but doesn't directly address it? These kinds of problems are generally the subject of data fusion.
  • A Bit of Hardware and a Bunch of Software: Though not always the case, our solutions tend to involve clever software with some custom or unusual hardware. Sometimes a laser makes its way in, sometimes a unique display, sometimes a microsensor.
  • Advanced Analytics: Most of our solutions have some math. Some of our solutions have a lot of math.
Here is some of our past research:
  1. SONUS: Self-organizing networks of unattended sensors.
  2. Deception Robust Control: Decision making and situational understanding in an environment with competitors or adversaries who employ deception, or in the face of self-deception.
  3. Strategic Network Discovery: Social network analytics for discovering groups of people who are participating in activities of interest such as money laundering, narcotics trafficking, or disease epidemics, etc.
  4. Distributed Fusion Management: Distributed level 1 data fusion in highly unstable low bandwidth networks.
  5. Stochastic Short Read Alignment: Extreme compression of genomic sequence data from next generation sequencing processes.
  6. Laser Wave Mixing: Using infrared lasers to discover minute traces of explosives, toxins, marker proteins, etc. at a distance.
  7. Probabilistic Logistics: Large-scale distribution networks where lots of variables are uncertain. Example, air traffic control.
  8. Personal Situational Awareness: Outfit a foot soldier, first responder, or law enforcement officer with a heads up display that marks objects of interest in the field-of-view and allows team members to mark those objects of interest with a gesturing glove.
  9. Vision as a Behavioral Process: Computer vision under a wide variety of scaled rotations and aspects.
  10. Hypothesis Management with Quantum Computers: Utilizing the new Adiabatic Quantum Computers (AQCs) that have just come into production, for solving complex problems in data fusion.

S3 Data Science
S3 Data Science, copyright 2015.