Industrial Research & Grant

Main project dashboard |full

Overview

Industrial research often aims to apply, validate, or extend existing academic research into specific sector use cases. This process can be more challenging because of noisy data and complex requirements from stakeholders with different perspectives. Research grant applications are often needed to support an industrial research program. Check out my GoodAI Research Grant project.

Challenge

The challenges in industrial research and grants are two-fold:

  • Research Proposal Writing and Interview: This involves preparing a strong proposal and a successful interview for an industrial grant application.
  • Ongoing Research Challenges: When a proposed method fails, it is crucial to stay on track with the original plan while finding alternative solutions.

Methodology

  1. Program Design: Identify the research goal and aim for potential final outcomes such as a patent, paper publication or application software. Propose methods to achieve the research goal, design proof-of-concepts, and evaluate performance to ensure the goals are met.
  2. Research Proposal: A clear presentation of the research program design, timeline, resources, and budget estimation in the research proposal.
  3. Project Management: Liaise with users, stakeholders, and grant providers. Provide periodic progress updates through written reports and meetings.
  4. Project Delivery: Follow typical processes in model training, patent filing, paper publication, and software development.

Tools

Collaborating with domain-specific experts is crucial in a research program. It’s also essential to use the right tech stack for experiments, development, and deployments. This includes: Python with PyTorch for development, AWS, Azure or GCP for cloud development, and GPU model training. Git and GitHub are used for version control and collaboration. Continuous integration/continuous deployment (CI/CD) practices assist with testing and deployment.

Results

Depending on the goals of the research program, the final outcome could be a patent, paper, or software. In the GoodAI industrial research grant, the result was a detailed internal report paper on the benefits of using a specific Bayesian framework for continual learning.