AI & ML Research 3 Days

Last updated: July 26, 2026, 11:30 PM ET

LLM Agents and Memory

An LLM agent using OpenAI's Agents SDK and Playwright MCP. To address limitations in context windows that prioritize recency over importance, a usage-reinforced decay engine was built for AI agent memory, drawing inspiration from the Ebbinghaus forgetting curve. Separately, prompting techniques were detailed to maximize developer efficiency.

Vector Search and Data Processing

When RAM costs become prohibitive, optimizing vector search, considering trade-offs between latency and storage for technologies like HNSW, SPANN, and Disk ANN. An intelligent document processing (IDP) system to automate the classification and extraction of PII from emails using AWS.

AI Foundational Models and Simulation

Tabular foundation models are introduced as a new class of models capable of predicting spreadsheet columns zero-shot, now outperforming tuned gradient-boosted trees on the Tab Arena benchmark. In a departure from traditional methods, a fluid simulator was developed that generates phenomena like the Kármán vortex street without directly solving fluid equations, utilizing the Lattice Boltzmann Method implemented in C++.

Biotechnology and Medical Innovations

Organ transplant efforts are seeing breakthroughs with the successful transplantation of supercooled kidneys into pigs. This research aims to address the critical shortage of donor organs, which is often exacerbated by the limited time organs remain viable outside the body.