Oddly, I in a previous life I was a product manager for an significant ML pipeline, and I can state with some confidence that the key to this effort as described will be someone to keep everyone focused on the objective and mitigate the tendency to bikeshed and yak shave. The risk is that the team loses its focus and individual engineers think achieving some novel result in the discipline will be the sufficient (and then, necessary) condition, where "if only we solve this ML problem I can coincidentally speak at conferences about, we will succeed." The other risk is where you get into a fundraising death spiral, where you can't produce or admit concrete results because you need to keep the ball in the air and hope alive to get your next round of funding. The way to avoid this is to have someone leading the effort who DGAF about social climbing with investors, particularly the kind of family money who will be drawn to this, imo.
I guarantee this project will not be solving new problems in ML, and everything they do will be implementing, scaling, and optimizing the compute required for existing methods. This is engineering problems applied to archeology, and not the need to solve computer/data science problems that require new science to achieve. Maaaybe you get some new IP for using ML to process lidar and gravimetry data (I know some people involved in doing this from space), but if I were pitching on this, I would lead with being open to new science, but demonstrate a track record on getting solved problems implemented. Make sure the incentives of your team are aligned and that they can commit to the mission, as side of the desk science projects are probably the main risk to this effort, I would speculate.
Comments
Oddly, I in a previous life I was a product manager for an significant ML pipeline, and I can state with some confidence that the key to this effort as described will be someone to keep everyone focused on the objective and mitigate the tendency to bikeshed and yak shave. The risk is that the team loses its focus and individual engineers think achieving some novel result in the discipline will be the sufficient (and then, necessary) condition, where "if only we solve this ML problem I can coincidentally speak at conferences about, we will succeed." The other risk is where you get into a fundraising death spiral, where you can't produce or admit concrete results because you need to keep the ball in the air and hope alive to get your next round of funding. The way to avoid this is to have someone leading the effort who DGAF about social climbing with investors, particularly the kind of family money who will be drawn to this, imo.
I guarantee this project will not be solving new problems in ML, and everything they do will be implementing, scaling, and optimizing the compute required for existing methods. This is engineering problems applied to archeology, and not the need to solve computer/data science problems that require new science to achieve. Maaaybe you get some new IP for using ML to process lidar and gravimetry data (I know some people involved in doing this from space), but if I were pitching on this, I would lead with being open to new science, but demonstrate a track record on getting solved problems implemented. Make sure the incentives of your team are aligned and that they can commit to the mission, as side of the desk science projects are probably the main risk to this effort, I would speculate.