Work
Production information systems and research prototypes spanning source-grounded AI, geospatial analysis, complex systems, and medical imaging.
AI reliability
A local, reproducible evaluation harness that decomposes RAG answers into atomic claims and compares NLI and LLM-judge assessments of source support.
Complex systems
Feb 2023 - Aug 2023
Agent-based simulations of collaboration networks in large software ecosystems, validated against Eclipse, Android, and OpenStack data.
Information systems
Dec 2023 - Present
A production retrieval-augmented document analysis system that produces structured, source-grounded summaries over heterogeneous document collections.
Information systems
Dec 2023 - Present
A configurable geospatial visualization tool for turning structured route and region data into interactive maps and exportable visual reports.
Machine learning
Mar 2023
A Streamlit application that compares resumes with job descriptions using contextual embeddings, cosine similarity, and skill classification.
Machine learning
Nov 2020 - Feb 2023
A research notebook collection spanning U-Nets, retinal and dendritic-spine segmentation, neural networks, sentiment analysis, transfer learning, and natural language generation.
Information systems
Dec 2023 - Present
A multilingual pipeline that collects and normalizes reporting and public information from regional web and feed sources for analysis.
Medical imaging
Jul 2022 - Oct 2022
Research pipelines for preparing CT perfusion data, denoising medical images, and estimating blood-flow parameters used to identify stroke regions.
Medical imaging
Jul 2021 - Jan 2023
A meta-consensus binarization and U-Net workflow for segmenting dendritic spine images, with graph-based measurements integrated into JYNIA.
Medical imaging
Apr 2021 - Jul 2021
Retinal vessel segmentation experiments using clustering, support-vector machines, neural networks, and class-balancing methods.
Machine learning
Aug 2022
An image-processing prototype that adapts clothing imagery between body structures using key-point detection and seam-carving techniques.
Machine learning
Theory-to-code studies of neural networks, continuous bag-of-words models, graph neural networks, and sentiment analysis.