AI Projects & Research

5-agent n8n workflow delivering a daily executive briefing from Slack, Jira, and Calendar — with confidential routing, orchestration-level graceful degradation, and error handling. 3-layer architecture: Control, Execution, Reliability.

8-agent adversarial pipeline: Greeter → Planner → dual Analyzers → Synthesizer → Critic → Reviser → Publisher. Routes any decision brief through structured debate to produce a paste-ready Decision Memo.

RAG-powered knowledge agent ingesting 7 sources into Pinecone — auto-generates Monthly Financial Standing Reports and drafts stakeholder emails from grounded vector retrieval.

Multi-agent voice + text Launch Command Center with 4 specialist agents (Technical, GTM, Executive) auto-routing stakeholder questions to the right domain via ROUTE: signal pattern.

No-code RAG financial intelligence system for 400 brokers — n8n workflows ingest news, stock prices, and SEC filings into Pinecone with LLM-as-a-Judge quality evaluation built in.

4-stage n8n workflow for ShopNest Global (50M+ customers, 200K+ orders/day) — summarizes raw support tickets, drafts policy-grounded customer responses, and scores both with LLM-as-a-Judge. Submitted grade: 60/60.

RAG-based policy Q&A for insurance sales reps on live calls — n8n workflow ingests 4 policy documents into Pinecone, answers grounded questions with GPT-4.1-mini, and scores every response with an LLM-as-a-Judge that hard-fails on hallucination.

3-agent Python pipeline automating Jira portfolio reporting — daily monitoring, escalation drafts, and a weekly 5-slide PPTX deck with Claude-generated analysis and executive narrative.

Retrieval agent for organizational knowledge across 8 Confluence spaces — connected via native MCP server for grounded, hallucination-free answers with CQL full-site search.