Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
Making local SLMs viable in agent loops: A hybrid approach with Jev
Published:
Decoupling discrete choice from text generation: a fast router picks the tool and target on each turn, so a local small language model only generates text when it is actually needed.
PMG: A Private Memory Graph for Persistent AI Memory
Published:
PMG encodes conversation history as a dual-topology graph — temporal chains within sessions, semantic edges across them — enabling structured, context-aware memory retrieval for AI assistants.
portfolio
MOF H₂ Storage Predictor: A Deep Learning–Driven Hydrogen Storage Prediction Tool
Published:
MOF H₂ Storage Predictor is a machine-learning–based tool designed to predict hydrogen storage capacity in metal–organic frameworks (MOFs).
SideHover — macOS Productivity App
Published:
A macOS productivity app on the Mac App Store: an edge-triggered floating browser with multi-tab browsing and notes.
c2d — Chat to Dataset
Published:
A multi-agent AI system for natural language data analysis. Upload a CSV or Excel file, ask questions in natural language, and get SQL queries, interactive charts, statistical tests, and a written report — all generated automatically.
SideCli — Auto-hide Sidebar Terminal for macOS
Published:
SideCli is a command-line tool built for a sidebar workflow, with edge auto-hide, instant pop-out, native shortcuts, and a minimal footprint.
asAgent — Observable Desktop AI Agent & Automation Runtime
Published:
Open-source desktop AI agent combining chat, browser assistance, scheduled tasks, and reversible file operations in one application.
MiniCPM5-2B-Jev — Calibrated Single-Pass Decision Model
Published:
MiniCPM5-2B adapted into a structured decision model that returns calibrated probabilities in one forward pass; #1 open-weight model ≤2B on JevBench.
publications
Multi-Level Structural Descriptors for Deep Learning–Based Prediction of Hydrogen Storage in Metal–Organic Frameworks
Published in Applied Sciences, 2026
A hybrid deep-learning framework using geometric, atomic, and crystallographic descriptors to predict hydrogen uptake in MOFs, used to screen 137,953 hypothetical MOFs and identify 630 candidates exceeding 50 g H2/L.
Recommended citation: Bai, Y., Aldrich, C., & Liu, X. (2026). "Multi-Level Structural Descriptors for Deep Learning–Based Prediction of Hydrogen Storage in Metal–Organic Frameworks." Applied Sciences, 16(19), 9574. https://doi.org/10.3390/app16199574
Download Paper | Download Bibtex