Agentic AI systems, automation pipelines, and apps built with LLMs and modern stacks. Everything here is real — no toy demos.
A multi-step autonomous research system. Input a topic, get a structured brief. No human-in-the-loop required for the research phase.
The pipeline receives a research topic, uses a Planner agent to decompose it into sub-queries, executes parallel web searches, retrieves and summarizes each source, then passes all results to a Synthesizer that produces a structured markdown brief with citations.
Built around Claude's tool-use API with streaming for real-time progress feedback. Designed to handle multi-hop research tasks that would take a human 2–3 hours in under 5 minutes.
Market scanner + AI context generator. Flags setups matching technical criteria, fetches options flow, and generates Claude-powered summaries per ticker.
Scans watchlist tickers for predefined chart patterns (breakouts, consolidations, volume anomalies). For each flagged setup, fetches options flow data and passes both to Claude to generate a concise context summary — bias, key levels, notable options positioning.
RAG-powered assistant for D365 FSCM configuration questions and X++ code generation.
Autonomous email triage — classify, prioritize, draft responses, surface action items.
Personal spaced-repetition system powered by AI that adapts to weak spots across D365, AI, and trading knowledge.