Elder Research’s R&D Drives Innovation and Expertise

by William Goodrum, Ph.D.

A 25-Year History of Innovation

Data Science and Machine Learning have been top topics in recent years in popular media and the private and public sectors, which Elder Research, as a pioneer in these fields, has welcomed. In the early years, it was rare to encounter people who fully appreciated the power of the technology. By focusing on delivering high-quality services and top value to our clients, rather than the latest technology itself, we built our firm’s reputation, as well as that of DS/AI/ML. Those who are new to Data Science may run into the hype and overinflated results that can result when Data Science isn’t applied with proper rigor. However, organizations can unlock real and sustainable value by properly harnessing this powerful technology to analyze their data and solve tough, but well-defined problems.

Innovation to Marketable Technology

Our success has not been limited to consulting services; we have also won multiple Small Business Innovation Research and Technology Transfer (SBIR/STTR) awards from Federal government agencies. Available only to small businesses, the awards are highly competitive, with an overall win rate of around 15%. The SBIR and STTR programs team innovative small businesses with government stakeholders to tackle some of the most challenging Science and Technology (S&T) and R&D problems that impede mission success. SBIR/STTR is focused on producing marketable technology derived from the R&D programs. Over the 40 years of the SBIR program, this has contributed to billions of dollars of economic output. Our SBIR R&D successes have been driven primarily through the work of our award-winning and highly-skilled software engineering team.

Why Elder Research Does R&D

R&D is unusual for a consultancy like Elder Research since our primary mission is not to develop and sell products. Still, there are strategic and important reasons for our R&D work:

• Staff Development and Entrepreneurship: Consulting projects have immediate implementation timelines and focus on proven methods, while R&D projects offer our staff the chance to work with more advanced techniques that are just coming out of academic environments. R&D also allows our staff to develop and test ideas on problems that have market potential and address real needs. This continuous learning benefits all of our clients by enabling our teams to develop more advanced solutions with greater success rates and delivering increased value.
• Stay Current in a Rapidly Evolving Field: Here are the results from the 2016 KDnuggets survey of Analytics/Data Science tools in use by Data Scientists:

And here is the equivalent top 10 list of Analytics, Data Science, and Machine Learning tools from 2019:

When comparing the two surveys, one important takeaway is that although there has been stability in terms of some top-level tools (e.g., Python, Excel), in just three years, Deep Learning tools like Keras and Tensorflow entered the Top 10 most used tools for Data Scientists. That is remarkable growth and a paradigmatic shift for Data Scientists and their field.

R&D provides our Data Scientists opportunities to work with cutting edge algorithms like Deep Learning; this enables our staff to develop the skills they need as individuals while also ensuring that our corporate service offerings are relevant and current. The Data Science horizon continues to change quickly, and the technology of tomorrow can very rapidly become the tool to solve the problem of today.

• Cut Through the Hype: There are some promising recent advancements in AI/ML, but there is also widespread hype and problems with replication of results. A recent article in Science News showed that claimed advancements in image recognition benchmarks were either overinflated or false. In aggregate, such bad work can undermine trust in applying AI/ML to solve business or mission problems. One of our corporate values is Integrity; we choose truth over expediency. In R&D, this means that we rigorously test and validate our results to ensure that the outcomes we achieve are accurate and reproducible. It also requires that we test new approaches and algorithms arising from academia to identify those with the most promising value for our clients.
• Synergistic Benefits with Existing Clients: Pursuing R&D through the SBIR and STTR programs allows us to tackle mission-critical problems for government and develop solutions that can be sold in the commercial market (a win-win for us and our customers). Demonstrating innovative work in SBIR/STTR and its relevance to the overall success of the mission reinforces our non-R&D engagements at our partner agencies.

How Elder Research Does R&D Better

R&D at Elder Research differs from that at a dedicated technology development firm or startup.

1. Agility + Experience: We are an established small business. Our business does not depend on developing or selling a specific technology. As a small business we are agile and innovative, but with the experience, history, and long view of an established firm providing successful Data Science solutions for decades.
2. Network Connections: Our long history and reputation in the Data Science field means that we have a large network of partners in industry, academia, and at prime contractors. We have won these connections, collaborators, and champions by repeatedly implementing AI/ML technology into real workflows and delivering measurable value.
3. Technical Depth: We have a deep and broad bench of technical talent spanning a wide array of disciplines. While we specialize specifically in AI/ML and Data Science, we apply an interdisciplinary approach to all of our projects. More than a dozen staff hold PhDs (and another two-score, Masters) in STEM disciplines. This in-house roster of diverse talent enables us to create lasting and impactful technology.

Our R&D Focus Areas

In our general consulting projects our AI/ML solutions span a vast array of problems and domains. In R&D we focus on the following AI/ML technology areas:

• Materials Engineering/Machine Reliability: Current adoption of AI/ML in Materials Science and Mechanical Engineering lags behind other fields for a variety of reasons. It is partly due to highly-developed and sophisticated techniques for solving problems related to predicting or forecasting machine and material failures. Such techniques are made possible by the proliferation of Internet-of-Things enabled monitoring and sensing devices. Our longstanding work in NASA STTR, in partnership with Southwest Research Institute, is the foundation for our success and leadership in this area.
• AI/ML Applications for Geospatial Analysis and Remote Sensing: The increasing availability of high-quality remote sensing data from satellite platforms, and decreasing costs of acquiring LIDAR data, has led to an explosion of available information about geophysical and atmospheric processes. Similarly, openly-available intelligence from overhead imaging has enabled the application of object detection and classification models to facilitate tasks in fields ranging from defense to humanitarian relief. Several of our promising R&D projects demonstrate novel applications of Deep Learning architectures for using these data sources to improve our understanding of the world.
• Text Analytics and Natural Language Processing (NLP): Elder Research has been a leader in Text Mining and the extraction of valuable insights from unstructured information throughout our consulting history. Recent advancements in NLP make it possible to automate the processing and knowledge discovery from text data sources. We have demonstrated the value of applying Deep Learning to identify risk features in text to our clients, and through R&D, we are testing the latest models from academia and accelerating the transition from the lab to practical use.
• Physics-Informed Machine Learning: Recent advancements in Deep Learning architectures combine the strengths of traditional physical models and nonparametric approaches. These new techniques meaningfully advance the understanding of physical processes such as turbulent fluid flow, the dynamics and control of autonomous vehicles, and numerical weather prediction. Our internal R&D efforts are leading the way to transform these game-changing innovations to practical, real-world solutions.

Summary

Our vision for Research & Development at ERI is to discover, incubate, and accelerate innovative and validated research into viable technology to solve very challenging data problems and address real needs. We have had the honor to work on many hard and rewarding DS/AI/ML problems in during our more than 25-year history. These past experiences form the foundation of our future R&D successes, which are the seeds of future value for our clients.

Originally published at https://www.elderresearch.com on February 5, 2021.