Systems Development at Discovery Logic

What is systems development like at Discovery Logic? How does it compare to elsewhere? To find out, in April 2007 we interviewed employees Marty, Laure, Matt, and Greg. Marty is a Senior Software Engineer. Laure is Science Director. Matt is Chief Technology Officer. Greg is Vice President of Research and Development.

What's it like to work at Discovery Logic?

Laure: It's fun. There is a lot of opportunity here to be creative and to collaborate. One of our core business functions is custom systems development. Typically, a client will say, "I'm having trouble doing X. Can you design a system that addresses this problem?" A DL team will brainstorm possible solutions and then discuss them with the client. We frequently create pilot systems to demonstrate our solutions, and these pilots often lead to larger projects. We then bring in others across the company for further development. As a scientist, such hypothesis development and experimentation is a real kick.

Greg: A big reason I'm here is that I like small companies. I like getting into small groups and growing things from there. Projects involve small teams (usually not more than eight people), so most people are able to work on a variety of tasks. You don't get pigeonholed into doing just one thing. And I like that we're able to stay on the front-end of technology. None of our technologies are more than a few years old.

Matt: Learning the latest technologies is fun, but our purpose for doing it is serious. We're trying to find innovative solutions for real problems, especially for NIH. I really admire its mission-saving human life. Anyone in the world can embrace that. Now, we are not directly saving human life, but we are facilitating others who are. Like Greg, I enjoy working in small to middle-sized companies where there is more flexibility in what the company can do. Small companies place more emphasis on satisfying the client, because we know that our entire livelihood depends on it.

Marty: I work on a project for the Center for Scientific Review (CSR) at NIH in Bethesda. CSR wants to shorten the time between when medical researchers submit grant applications and when they learn if their research will be funded. To help CSR, I and four other engineers created the proof-of-concept Automated Referral Workflow System (ARWS). ARWS speeds grant review by reducing the time to route incoming electronically received grant applications to appropriate scientific peer review committees. This is a huge task. CSR has over 200 peer review committees (called Study Sections) and handles more than 80,000 grant applications a year. Before ARWS, CSR assigned grant applications to committees manually!

I handle the full life cycle for the project: analysis, design, coding, testing, deployment, and operational support. It's been challenging and rewarding because the problem we had to solve was so new to me. ARWS uses text mining technologies in order to automate decisions on which committee is best suited to review each grant application. CSR is invested in the system and very pleased with Discovery Logic's work. I really enjoy supporting their mission and making their job easier.

How is Discovery Logic different from other places where you've worked?

Laure: It is a cliché, but it's apt: Discovery Logic is agile. A client can describe their business problem, we can discuss possible solutions, and DL can submit a proposal all in about a week. Sometimes in two days. We can deliver a pilot project in as little as 4 weeks. I've worked in places where just the proposal process would take six months to a year. By the time you get approval, you're either no longer interested, or six other people have jumped on it. Because DL is willing to invest in experiments, we can take advantage of opportunities very quickly.

Greg: What really stands out compared to other companies where I've worked is the commitment to get the best. Management doesn't cut corners with the quality of people we hire. Compensation is pretty strong so we can attract great people and build great systems. One of the comments I heard from two employees we hired in the last five months is that they were surprised by how smart everybody else was. They weren't expecting that. There are a lot places, particularly among government contractors, where it's just bodies, taking anyone you can get. We don't go for that. I hate hiring below A-class employees. I've never found it worth it. Management has the same commitment.

Marty: I agree, salaries here are very competitive. Management does not hesitate to buy tools and equipment when they will help make the engineers more efficient. No other place where I've worked has been as willing to spend money in order get a highly qualified and productive engineering team.

How do you work with clients?

Laure: We're known for custom software development. Sometimes clients find that software packages from big firms like IBM don't really meet their needs. No amount of shoehorning clients into these preconceived solutions gives the results they want. What we do is like creating shoes that fit perfectly. We have an excellent reputation as a company willing to do experiments and work closely with clients to solve problems.

Greg: I often compare our initial contact with clients to taste testing at Baskin-Robbins. Once you taste an ice cream you like, you're much more likely to buy. With custom systems development, it's similar. Clients like to have a taste of what we can do before signing up for a full-blown project. So sometimes we demonstrate our capabilities by developing proof-of-concept or prototype systems for the client. When our prototypes provide better results faster than they've been getting, they're sold.

Matt: This is one of the things I really like, that senior management is willing to invest in prototype technology development in order to show what we can do. In tech companies I've worked at before it would be difficult to get a commitment to invest as quickly as we do here. Sometimes clients have short time frames for solving problems, so investment decisions have to be quick. Because the owners of Discovery Logic are so involved in the technology they've been able to see these projects as fantastic opportunities.

Greg: We usually run the whole project for the client, from beginning to end. After they state what the problem is, we'll figure out the solution, implement it, deploy it, support it, and in some cases provide staff in the client's office to use the system. We provide a complete solution. That's more valuable to clients, and our ongoing involvement in their projects gives us more stability as a business.

Matt: One of my jobs is to insure that we have really good technical teams for these projects. We want every employee to be working on challenging and interesting problems. I try to make sure that employees know about upcoming projects so that they have the chance to stretch their skills. If they've been working on something for a long time, I try to let them do something else. I think we're doing a reasonably good job at this. Turnover at Discovery Logic is very low.

Do you have a software development specialty?

Matt: A lot of companies and government agencies need software to automate business processes that are fairly tedious. And, many of our technology solutions have been developed for this purpose. Sometimes the goal is to develop a text mining system like ARWS that mostly replaces human decisions within the business process. At other times our objective is to improve the information available to decision-makers and greatly speed the process.

Some of our most interesting process automation projects call upon our special capabilities for information fusion. By this we mean linking together multiple disparate databases as if they were a single database, so that information otherwise extremely time-consuming to obtain can be retrieved quickly.

Greg: There are a number of names for this besides information fusion. Some call it data mining, textual pattern recognition, text mining, entity extraction, or knowledge management. Whatever you call it, linking multiple databases involves finding relationships between data sources that are not explicitly provided. It is easy, for instance, to link two databases that store social security numbers. Those numbers explicitly relate the records the databases store. But if data sources have unstructured data or free-form text (for example, news articles, research grant proposals, patents, or progress reports) finding data relationships is a much greater challenge.

A lot of academic research has been done on these problems and we have made a specialty of using this research to solve day-to-day problems for our clients. We can find relationships that allow data from diverse information sources to be pulled together. And when we create systems that make searching disparate databases easy, our clients are able to see relationships in data they've never been able to see before and make better decisions as a result.

Matt: To give a brief example, we recently helped the National Institute of Allergy and Infectious Diseases (NIAID) come up with a system for tracking outcomes of research projects for which grants have been awarded. The division's approach to evaluating research project outcomes was a time-consuming, tedious manual process and it is not surprising that it was rarely used. Consequently, some grant funding decisions were based on little more than "gut instinct." We were challenged to create a more efficient and consistent means for providing quantifiable research outcomes information through whatever means we could come up with.

After conducting a survey of published work on this topic, we expanded on paper published by researchers from Sandia National Laboratories and Indiana University to quickly build for NIAID a proof-of-concept information fusion system that links grants to numerous research outcome indicators. These include scientific publications, impact factors, citation counts, co-author networks, invention reports, patents, FDA-approved products, and Phase III clinical trials, among others. Our prototype system was well received and demonstrated how we could give our client new insight into research productivity from funded grants. This work has led to implementation of a production system that will be able to analyze grants on a much broader basis.

Any last words for systems developers interested in working for Discovery Logic?

Laure: We value people who are creative, like to work with others, and who adapt to new situations and new technologies. We may hire an engineer or manager for a particular project, but we are always looking beyond a single project.

Greg: We stay current with new and exciting technology. And we really make an effort to bring in the best people. A lot of people say that, but we really do. As a result, we have a tremendously successful track record and a great team.

Matt: At any company, the people you work with are as important as anything else. Bright people like to work with other bright people. You'll find that here.