groundedness

Provider-agnostic enterprise RAG and agent evaluation harness for Azure Foundry, vLLM, Ollama, and local demos.

Evaluation patterns, release gates, and anti-hallucination techniques for developer-focused AI workflows.

Single-agent, evidence-grounded claim verification to catch LLM hallucinations — a pluggable fact-gate for agent-arena and any multi-agent system (CrewAI, AutoGen, LangGraph).

Bilingual reduced-fare policy assistant for six California transit agencies, grounded in dated citations and shipped with a public 201-case evaluation plus independent black-box audit.

Scores whether an AI agent's answer is actually supported by the tool results it had access to. An LLM-as-judge harness for catching ungrounded or fabricated claims in agent outputs.

TypeScript eval harness for measuring whether Grok answers stay grounded in source evidence

Detect & score LLM hallucinations by groundedness — labeled data, precision/recall/F1, runs offline with no API key. Pluggable LLM-judge backends.

Deterministic citation and claim-support checks for RAG evaluation datasets.

RAG that grades every answer: groundedness, helpfulness, accuracy + citations. FastAPI + live browser demo

RAG evaluation workbench for retrieval recall, citation coverage, groundedness checks, and failure analysis

Your RAG, put to proof — self-correcting answers and CI eval gates that block the merge when hallucination rate crosses your threshold

A local groundedness judge for RAG: QLoRA-distilled to match a frontier judge 100% at $0/call. Ships only if its own evals beat baseline.

What the four-box RAG diagram leaves out hybrid retrieval, chunking, security trimming, and groundedness evaluation, as runnable Azure AI Search samples.

Groundedness & hallucination checker for LLM/RAG answers: decomposes an answer into atomic claims, verifies each against its sources (LLM judge or offline NLI), and highlights what isn't supported.

Bilingual (Arabic/English) medical RAG system with a runtime groundedness gate that blocks unsupported answers before they reach the user.