Reimagining innovation with cloud migration and machine learning solutions for an eminent US-based market research provider.

Elevating the script with NLG and machine learning

Survey writers took an average of 10 minutes to write a question and spent most of the time generating false answers for the questions. The incorrect answers needed to be grammatically similar to the correct answer but wrong enough to ensure clarity among survey takers. We built a tool that analyzed the writers’ questions and could generate the kinds of false solutions they otherwise would need to think of and write out. The tool would generate answers, and the writers could select their preferences. If none of the answers were appropriate, they could generate new suggestions. Whatever they chose was fed back into the tool, so it constantly improved its recommendations. Ultimately, our tool developed incorrect answers instantly, reducing the average time to write a question to 6.7 minutes and saving nearly 50% of the question-generation time. It helped support newer writers as they learned from the tool-generated answers instead of relying on more experienced writers. Our dedicated efforts ensured the tool was awarded ‘Best Innovation in NLP’ in the AIconics awards at the AI Summit, New York.

Bringing efficiency and quality control to data migration

Our client sought cloud migration support and wanted to reimagine innovation, and that’s where we came into the picture. Cloud migration becomes cumbersome to manage over time. On the other hand, with FitNesse, we could test sample data during the data migration. Whereas other cloud migration strategies require engineers to test each piece of data individually or risk migration of corrupted files. We built a test suite with FitNesse that tested sample files and data. Now, we could also assume that similar files and data migrated successfully if they migrated without incident. The result: We successfully migrated files without taking all the time to validate each one. Our app highlighted the files that failed. It addressed them and all similar files, so engineers only had to focus on issues and not worry about successful file migrations. Even better, the tool recorded all results in Excel spreadsheets and on the FitNesse wiki to ensure quality remained high and that almost anyone with permission could access test results.

Building a firm foundation with Software Development

Many aspects of our work with the client involved innovation and bringing new ideas to well-established processes and markets, so we knew we would need to build software engineering practices into each part of the project. To do this, we put together a team with a robust, strategic foundation in software product development. We began by defining the support processes. We then recommended and configured Jira Service Desk to manage tickets. We used Java and Angular for application development and enhancement and Oracle PS/SQL for app support. When we supported their migration to AWS Cloud, we ran database administration through Oracle and used Manual, Selenium, and FitNesse for quality assurance. When we developed the tool to generate answers for the survey writers, we utilized Terraform scripting to configure the automation. We integrated the tool with Google Docs and a video player so writers could craft questions and answers all in one place without the need to open and close many different apps and windows.