ML Engineer — Wildlife Population Analytics

Zara Patel

Part-time · Mid-level · Johannesburg

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We are a wildlife conservation organisation. We work across sub-Saharan Africa on the protection of large mammal populations — elephant, lion, wild dog, and cheetah — in collaboration with national park authorities, local communities, and other conservation partners. Our technical team uses camera trap networks, GPS collar telemetry, and aerial survey data to monitor population dynamics and detect human-wildlife conflict events before they escalate. We have more data than we can currently analyse. Camera traps generate between 80,000 and 120,000 images per month across our monitoring sites. We currently rely on volunteers and interns to classify those images — identifying species, counting individuals, and flagging conflict-relevant events. It is slow, it is expensive in volunteer time, and it introduces classification inconsistency. We are hiring a part-time ML engineer — two days a week — to help us build automated image classification systems that can handle the routine species identification at scale, so that our ecological team can focus on the analysis that requires expert judgement rather than counting zebras in photographs. We are a charity. Our budget is constrained. We are offering what we can for work that we believe is genuinely worth doing.

Responsibilities

Build an automated species classification system for camera trap images across our monitoring network
Develop an active learning workflow that prioritises difficult or uncertain classifications for volunteer review rather than blanket manual labelling
Build evaluation metrics that make sense for ecological monitoring — species recall rates by site and season, not just aggregate accuracy
Collaborate with our field ecologists to define what the system needs to get right versus what it can afford to get wrong
Document everything clearly enough that the next person who maintains this — likely not an ML engineer — can understand it

Requirements

3+ years of ML engineering with hands-on computer vision experience — experience with wildlife camera trap data, citizen science image datasets, or ecological monitoring applications would be directly valuable
PyTorch for developing image classification models that work on small, unbalanced, domain-specific datasets rather than benchmark-scale data
Python for pipeline development, data cleaning, and building the tools our ecological team can use without ML expertise
Scikit-learn for evaluation and the baseline models that give us something to compare against
MLflow for experiment tracking — with limited compute and limited labelled data, every experiment needs to be reproducible
SQL for working with our sighting database and camera trap deployment records
A genuine interest in wildlife conservation — we are not a technology company, we are a conservation organisation, and the fit matters

Benefits

£28,000 – £36,000 pro-rated to two days per week
Fully remote — anywhere with reliable internet
Flexible hours — we care about the output, not the calendar
Opportunity to visit monitoring sites in southern Africa if you would like to understand the field context — we cannot pay for this but we can arrange access
Work that is genuinely about something — the elephant population we are monitoring has declined 40% in the past decade, and the data we collect informs policy decisions that could change that