Motahare Shahabi

Led a 9-member AI × Energy task-force that designed three proof-of-concept modules for a cloud-based virtual power plant (VPP) aggregating residential PV + battery assets (≈ 5 MW). Modelled household energy-consumption behaviour using smart-meter, weather, and occupancy data; demand-response algorithm cut peak load 18 % across a 120-home pilot. Secured €60 k seed funding and negotiated an MoU with Tehran Regional Electric Co. to scale the pilot to 1 000 homes in summer 2025.

Delivered a 12-week cognitive-rehabilitation programme for out-patients with schizophrenia (≈ 40 participants); average working-memory z-score rose +0.6 and processing-speed improved 15 % (p < .01). Designed gamified, tablet-based tasks targeting executive function; session-to-session adherence reached 87 % across 36 scheduled sessions. Curated a multi-modal psychiatric dataset (neuropsych tests + behavioral metrics + EEG) spanning schizophrenia, bipolar disorder, and MDD (total n ≈ 120).

First-author paper: conceived and implemented a dual-branch CNN (depth-wise + standard kernels) that fuses digitised hand-drawing trajectories with transformed EEG spectrograms to classify Alzheimer’s disease stage; manuscript now under submission to IEEE JBHI  Master’s-thesis spin-off: built multimodal progression-prediction models that integrate EEG biomarkers with Montreal Cognitive Assessment scores for 180 participants, generating personalised risk curves that guide the lab’s current longitudinal studies  Systematic review & meta-analysis (co-author): mapped >40 transformer-CNN-attention architectures for dementia detection, providing the first pooled effect-size estimates of EEG-based models (Elsevier journal, under review)  Statistical lead: applied Bayesian mixed-effects models to four-year MoCA trajectories, revealing three cognitive-decline phenotypes that informed the inclusion criteria of an ongoing interventional trial. Data-pipeline steward: established SOPs for recording, anonymising, and REDCap-ing multimodal data (EEG, handwriting, speech) — now the lab’s reference workflow. Mentorship: supervised five MSc students on EEG preprocessing and scikit-learn best practices; three have since presented at IEEE conferences.

• Used random-forest & LSTM on multispectral + human-activity data → predicted urban heat-stress zones (R² = 0.83). • Designed & delivered GIS labs for 120 undergrads across 3 semesters.