Search the Corpus

An AI career coaching platform that uses multi-agent orchestration and RAG, plus continuous evaluation using an LLM-as-judge mechanism.

Provide a Graph Neural Network (GNN) surrogate model to emulate blast-wave propagation and related outputs (e.g., pressure/impulse fields) for large-scale explosion simulations, potentially including hybrid coupling with LS-DYNA.

Demonstrates and provides methods (“trace inversion”/synthetic reasoning reconstruction) to recover detailed reasoning traces from LLMs using only observable artifacts such as the input, final answer, and optionally brief reasoning summaries—despite the target model withholding full chains of thought.

Desktop AI agent (“Aion Bot”) that uses natural-language instructions to control a user’s computer, coordinating with multiple LLM backends (Claude/GPT/Gemini and local via Ollama) and driving UI/browser actions via Playwright.

Framework (and associated method) for end-to-end autonomous driving using vision-language-action (VLA) models with a reactive, multi-agent latent-space rollout/supervision approach to improve closed-loop robustness in dynamic driving scenarios.