Open-source verifier for document citations, source evidence, RAG, and agents, with built-in document parsing.

Agentic Legal RAG / MCP Harness for Taiwan Law.

FastAPI service that checks extracted knowledge graphs on two axes: deterministic SHACL conformance (keyless) and LLM source-grounding (BYOK). Azure Container Apps, OIDC CI/CD, OpenTelemetry.

Production-oriented template for building AI agent skills as verifiable software components — offline eval harness, source-grounding validators, structured logs/traces/metrics, replay artifacts, and a CI quality gate. Runs fully offline with a deterministic mock model.

Synthetic SMB operations eval for source-grounded, privacy-safe, human-approved business agents.

Local-first AI job discovery workflow with source-health checks, job-card matching, and human-review application gates.

High-precision information extraction framework.

Full-stack AI financial research assistant with FastAPI, React, local RAG, source citations, portfolio risk summaries, and a deterministic RAG evaluation harness.

Eval-first RAG that checks claims against arXiv papers; cites the spans it uses, abstains when the corpus is silent, and measures every component against a frozen, passage-anchored gold set.

A standalone MCP suppressor that enforces grounding by validating references, rejecting unsupported claims, and stabilising model outputs.

Leadership notes on trusted financial AI, platform strategy, and governed agentic systems.

Public product and architecture brief for a local-first AI advisor workbench.

A source-grounded, multi-tradition debate council for Claude Code, Codex, and MCP—with structured claims, citation verification, fail-closed boundaries, and traceable authority rendering.

Trusted financial AI, agentic workflows, and senior AI product/platform architecture.

AI-powered RSS digest for 90 high-signal technology blogs, with Gemini scoring, category grouping, trend highlights, and Markdown reports.

Evidence-aware editorial AI workflow that turns topics into source-grounded draft packages using query contracts, source scoring, candidate validation, claim checks, and copy-readiness gates.

AI marketplace assistant demo with grounded recommendations, source-backed answers, and embeddable chat widget preview.

Public-safe research-to-deck workflow with deck contracts, HTML preview, QC reports, and traceable run bundles.

A full-stack AI-powered platform for verifying whether generated claims are supported by source documents.

Architecture notes on provider-aware financial data, advisor workflows, and trusted wealthtech AI.