OceanLum applies proprietary deep learning to a problem researchers have wrestled with for decades: identifying individual marine animals from amateur underwater footage at commercial scale. Our custom-trained neural networks recognise specific animals — not just species — by their unique biometric markers: spot patterns, scarring, fin shapes. Each sighting is geo-tagged and linked to its source excursion, generating a verifiable encounter history per operator and a longitudinal record per animal. The pipeline handles the conditions that break general-purpose vision models: variable lighting, motion blur, partial occlusion, and the wide gamut of consumer capture devices.