PostgreSQL extensions managed as stackable Docker layers.

Robust statistical AI-generated text detection framework.

Budget-adaptive routing implementation for optimizing edge-cloud inference latency and accuracy.

Code-mixed benchmark for evaluating LLMs on Romanized Indic-English instructions.

A frontend runtime for your app's dependency graph.

A framework for efficient GenAI-driven avatar tutors with local-first execution on consumer-grade hardware.

An AI-powered code editor used by a significant portion of Rivian's software team.

A verifier-guided repair framework for generating verified code in Dafny.

Modality-Agnostic Refusal Steering for training-free multimodal safety in LLMs.

A framework for improving medical visual question answering through evolving principle-guided supervision.

Self-correcting Autonomous Grounded Experimenter for reliable scientific research agents.

An LLM-based multi-agent workflow for refactoring software into HLS-compatible programs.

A language-to-execution safety suite for grounding lab rules into runtime guards for AI agents.

A suite of simulation environments probing LLM belief trajectories under multi-turn evidence accumulation.

Benchmarks (AIOps2025, RCA100) for evaluating LLM agents in microservice failure diagnosis.

A framework for semantic-aware, physics-informed, and geometry-grounded weather video synthesis.

High-performance inference engine implementing HARD-KV for efficient KV cache compression.

Modular framework for 3D MF-EIT stroke reconstruction including forward solvers and D-bar implementation.

A five-stage PDP/PEP for agent tool calls to provide deterministic value authorization gating.

A benchmark for evaluating temporal visual reasoning in video-to-code generation.