A student who loves sharks and knows how to code is hitting a wall, and that wall is a lack of validated, accessible data for conservation. This is a familiar frustration, but it is also a sign of a deeper problem in ocean science: data exists, but it is not yet integrated. The student's instinct to model shark populations from bycatch records, rather than rehashing attack statistics, is exactly the kind of purpose-driven thinking that moves us from anecdote to empirical understanding. At World Data Ocean, we see this as a call to build the integrated data ecosystem that makes such work possible, not just for a capstone, but for the entire research community.
The student is right to be wary of building a dataset from scratch. Hand-curating NOAA fishery reports is labor-intensive and introduces variability that is hard to calibrate. Yet their dilemma points to a critical gap. We have published guidance on Two career pathways to ocean intelligence: aquatic technician or fish culturist, which highlights how technical skills are merging with traditional marine roles. A computer science student who can wrangle messy catch logs into a clean, machine-learning-ready format is performing a function that sits squarely between those two pathways. Similarly, the precision required to identify a larval sea star's tube feet, as detailed in Tracking Development: Identifying a Larval Sea Star's Tube Feet, is the same kind of meticulous observation that turns raw bycatch reports into longitudinal, peer-reviewed indicators of population health. The student's work is a natural extension of that tradition, scaled by computation.
Our take is straightforward: this student should not abandon the project, but they should also not try to do it alone. The difficulty of finding a ready-made dataset on shark finning or bycatch rates is not a failure of the student; it is a failure of data accessibility. The solution is to treat this as a collaborative problem. Reach out to the fisheries scientists who compiled the NOAA reports. Ask them what variables they consider validated and which are still uncertain. A model built on a small, high-quality, calibrated dataset is more useful than a model built on a large, noisy one. The student's worry about creating their own data is valid, but a hybrid approach, using existing peer-reviewed catch data, then applying machine learning to estimate population trends, is the most practical path forward.
The specific consequence we want readers to watch for is this: if the student succeeds in building a predictive model from publicly available bycatch records, they will have created a replicable method. That method, once documented and shared, becomes part of the global ocean intelligence infrastructure. It is not a grand announcement or a flashy product. It is a measurable, real-time tool for conservation. That is the kind of impact that matters.