Nurturing tomorrow’s problem-solvers sits at the heart of our purpose. Every Hamoye intern is immersed in real work from day one, translating classroom theory into live, production-grade tasks. Throughout the programme they tackle end-to-end machine-learning projects, a process that both deepens their conceptual grasp and hones the practical instincts required to solve genuine business challenges. In the words of Albert Einstein, “Learning is experience. Everything else is just information.”

HDSC-OAU Spring ’ 25 Capstone Projects

Details of the premiere projects are provided here. Find your project details using the project topic assigned to your project group.

Deep Learning

Topic: CAI-Driven Battery Health & Life-Cycle Predictor (Dataset- Various- See Slack channel)

Project Instructions

Using the open lithium-ion battery datasets listed in the brief, carry out a streamlined study that ingests and cleans the cycle data, extracts a compact set of degradation features, trains at least one classical regression baseline alongside one sequence-aware deep-learning model to predict State-of-Health and Remaining Useful Life, and distills the findings into clear accuracy metrics, trend visualizations, and actionable guidance for extending battery service life and informing second-use decisions.

LSTM

Topic: Predictive Maintenance for Industrial Equipment (Dataset- Various- See Slack channel)

Project Instructions

Using the datasets provided, NASA’s C-MAPSS turbofan run-to-failure dataset, carry out a concise end-to-end investigation: first chart key sensor patterns across each engine’s life cycle, then create a compact set of time- and frequency-domain features that capture emerging degradation, next train and compare at least one classical regression baseline and one sequence-aware deep-learning model to forecast Remaining Useful Life (RUL), and finally distill the results into clear trends, forecast accuracy metrics, and practical maintenance recommendations that demonstrate how data-driven prediction can minimize unplanned downtime and cost.

Random Forest

Topic: Data-Driven Reservoir Characterization Using the Volve Field Data (Dataset- Various- See Slack Channel)

Project Instructions

Using Equinor’s open Volve North Sea field dataset, carry out a streamlined end-to-end workflow: first download and organise the well-log, seismic and production files, then visualise key subsurface curves and production trends; next engineer a concise set of spatial and statistical features and apply geostatistical interpolation (e.g., ordinary kriging) alongside at least one tree-based machine-learning model to map porosity, permeability and forecast well-level output, benchmarking each approach with spatial cross-validation; finally transform the best models into 3-D property grids and a short set of field-development recommendations that illustrate how data-driven reservoir characterisation can guide infill-well placement and boost recovery.