Compression methods and systematic review.
Two active research tracks alongside the client and product work: a model compression method, and a co-authored umbrella review on AI in energy management.
A neural network compression method.
Semantic Neuron Consolidation identifies semantically similar neurons within a trained network, groups them by functional similarity, and consolidates redundant units to shrink the model while preserving performance. The work is being prepared for a top-tier venue — NeurIPS, ICML, ICLR, AAAI, KDD or TPAMI/JMLR — with multi-seed runs, statistical significance testing, fair baselines and full ablations.
Mapping the Meta-Landscape of Artificial Intelligence in Energy Management: An Umbrella Review.
Role: co-author, Reviewer 1, data and pipeline coordinator. Methodology: PRISMA 2020 and PRIOR reporting guidelines, AMSTAR 2 quality appraisal, corrected covered area (CCA) overlap analysis, and a dual-track narrative plus bibliometric synthesis.
- Records identified
- 10,206Scopus · IEEE Xplore · OpenAlex · Lens
- Duplicates removed
- 5,174
- Unique records screened
- 5,032Dual-blinded human screening
- Search-string recall
- 81.8%Against a 22-paper gold-standard seed set
Have a system that needs to work in production?
Tell me what's breaking — or what you're building.