Research

The architecture of edge intelligence

Capicú researches how biological representations and efficient AI can deliver analytics, summarization, and interpretation directly at the instrument.

Focus areas

Our work spans edge computing, bioinformatics, process modeling, efficient inference, and evidence-grounded biological foundation models.

Publications

1. Retention-Constrained Post-Training Quantization of Cellpose-SAM for Stem Cell Microscopy

   Cruz Romero, S.A. · NeurIPS 2026 LatinX in AI (LXAI) Workshop · 2026

2. Proxy Metrics Can Hide Downstream Collapse: A Case Study on Compressed Cellular Segmentation

   Cruz Romero, S.A. · NeurIPS 2026 Workshop: I Can't Believe It's Not Better (ICBINB) · 2026

3. Tracking Scientific Information Loss in LLM-Assisted Biomanufacturing Process Development

   Cruz Romero, S.A. · NeurIPS 2026 Workshop: New in ML · 2026

4. INT8 Quantization Makes ARM Edge Inference Dispatch-Invariant

   Cruz Romero, S.A.; Maldonado Guerra, S.E. · arXiv preprint arXiv:2607.23227 · 2026

5. A Remote Smart Health Framework for Anemia Risk Stratification via Edge Medical Vision Systems

   Cruz Romero, S.A.; Mercado Hernández, M.J.; Ali Rivera, S.Y.; Santiago Fernández, J.A.; Lugo Beauchamp, W.E. · Applied Sciences, 16, 4924 · 2026