Proposed Research Plan

My purpose over the next five years is to improve execution of complex collaborative processes such as registering businesses, managing contracts, executing real estate transactions, facilitating negotiations, and enhancing patient care by focusing on supporting front-line decision makers with intelligent data systems. These systems are designed to adapt to real requirements and problems, rather than only plan for an ideal outcome or track deviation from the plan. In this way, contract requirements can be extracted and met, complicated and changing rules can be understood and addressed, emotionally-charged disagreements can be streamlined, and effective accounting and reporting automated. In a sophisticated civil society, these types of bureaucratic complications will only get worse until we find ways to address and streamline them.

To my knowledge, no current data solutions address this problem in a manner that remains flexible, comprehensible, and adaptable for any significant period of time. Rather, these systems devolve into control systems. Processes and expectations harden around old habits. Systems like ERP solutions, company policies, even workflow tracking with a spreadsheet lose their effectiveness because they focus work on the process rather than the outcome. With intelligent systems, processes that must remain rigid are held rigid, but every other system, especially for problem solvers focused on finishing work effectively, we can learn to adapt. This includes extracting the requirements out of a contract so that rules can be followed, escalating a communication problem so work cycles are not lost, tracking time, expense, and budget and leaving proof that this accurate and complete.

This also means learning how language unveils the intention, requirements, and priorities of an agreement. By directing this system towards complex, bureaucratic problems, I can iteratively improve my ability to solve these kinds of problems.

Currently, I have a working system for processing events on an event log, storing data, interpreting workflows, communicating relevant information to stakeholders, and storing information around agreements. This system is being applied as products in 3 markets that will be launched in the last quarter of 2019.

My next research steps involve mastering enough machine learning and natural language processing skills to deploy better tools around the contracts, conversations, and rules that come up when solving these problems. By combining smart machine learning features with smart workflow systems, we can incrementally improve how we collaborate without trying to reduce all of human interaction to an algorithm.

Lab work will primarily be started in Jupyter notebooks. Versions on datasets, models, notebooks, and libraries will make my research reproducible, more effective than anything I’ve seen in research labs over the last decade. Careful interfaces, tests, and deployment scripts will make my research many times easier to deploy and maintain in production. Integrating this research with workflows creates an adaptive and efficient solution to many complex problems.

Specific natural language processing tasks that I am taking on in the last quarter of 2019 are text classification, sentiment analysis, named entity recognition, models using sparse data, dialog agents, word meaning dependency, machine translation, question and answer, and text summarization. All of these tasks will have working solutions in a library, but will also be implemented in workflows.

Beyond research for my own projects, I am available to consult with other organizations in setting up their research labs, hiring a data science team, deploying their models, or using the latest research in their work. As I complete and publish more work, it will become more evident which advantages companies gain from my experience and expertise.