Let's (re)define the team structure:

=====================

Science Subsystem

all sub teams w/in science with minimal non-overlapping members with a dedicated sub-lead.

• Unmixing unmixing w/ statistical methods requirements design for payload specs developed great unmixing model yielding good results on lab dataset For the FINCH EYE dataset, anticipate a lot of additional work given (b/c data pipelines break, slow iterative process that involves "guessing" imaging conditions which yield the most accurate results). Keep senior members to develop more advanced unmixing methods, while training new members to do the remaining FINCH EYE work
• Deep learning synthesize hyperspectral data unmixing w/ deep learning models work mostly finished, will continue doing complimentary research (EO Deep Learning)
• atmospheric modelling pipeline mostly automated, running out of FINCH EYE work to do input optical properties of scene objects & imaging conditions -> BRDF -> ToA Radiance post-processing apply sensor response future work: add features to the pipeline canopy effects, man-made objects, etc.
• PAY Demos: very easy, can do as new member onboarding project

=====================

Mission Research Team

two members right now, no lead yet
dedicated entirely to PAY projects
educational space telescope

=====================

Data Processing Subsystem

• de-striping / de-noising (# members
• super-resolution
• calibration & verification

One big ML team but several projects running concurrently. Members contribute to any project. Some learning about "classic" sensor calibration methods, on hold since progress stopped om FINCH.

For the short term, 50-50 split between ML research & PAY projects. keep ~4 members + unused surplus from PAY projects for ML research.

=====================================================

Regarding a separate PI for Science & Data Processing

Difficult to find Profs doing exactly this, instead find Profs working on adjacent disciplines (e.g. ML specialist & Earth Science specialist). Don't want to become an extension of the Prof's research group. Also useful to apply for computational grants, which are needed for serious ML research work.

You have my permission to email profs & industry partners to get advising autonomously. If you can cc or bcc me & Mario that would be good, but I won't interfere with your process.

PAY PROJECTS FOR SCIENCE & DATA PROCESSING (Science & Software Data Processing)