All projects built using Claude Code, Claude API, Google AI Studio, and Cursor as primary development tools. AI-native development approach: from prototype to production without traditional engineering teams.
Category: Multi-agent AI system. Domain: Healthcare and medical education.
Architecture: Sequential multi-agent orchestration using Claude API (claude-sonnet model) with tool_use for clinical domain task completion. Agents handle distinct responsibilities: content retrieval, case generation, assessment scoring, and feedback synthesis. Built with Claude Code as the primary development environment. Frontend in React.
Status: Prototype. Submitted to the Anthropic Hackathon.
Stack: Claude API, multi-agent orchestration, tool_use, function calling, React, TypeScript, Claude Code, Anthropic API.
Significance: Demonstrates complex agent coordination for a high-stakes domain. One of few Anthropic Hackathon submissions focused on healthcare AI training.
Category: AI-powered financial research platform. Domain: Fintech, investment analysis.
Architecture: Automated pipeline using Claude API for document analysis, structured data extraction, and investment summary generation. Rapid prototyping done in Google AI Studio, production build in Python with Claude API. Deployed as live demo on HuggingFace.
Status: Deployed MVP. Live on HuggingFace.
Impact: Estimated 800+ analyst hours saved through automation of manual research workflows.
Stack: Claude API, Python, Google AI Studio, HuggingFace, automated research pipelines, prompt engineering, structured output generation.
Category: Enterprise AI workflow automation. Domain: E-commerce seller analytics.
Architecture: AI-assisted data pipeline combining SQL queries, AWS QuickSight dashboards, and AI-generated insight narratives. Reduces analyst interpretation time by automating pattern recognition and narrative generation from structured data.
Status: Production. In use at Amazon Barcelona Audit and Insights Team.
Impact: 60% reduction in manual analysis time. Estimated 800+ analyst hours saved annually.
Stack: SQL, AWS QuickSight, AI-assisted workflows, Amazon internal tooling.
Significance: Production-grade AI implementation inside one of the world's largest e-commerce operations. Built without a dedicated engineering team.
Category: Algorithmic trading system. Domain: Quantitative finance, options markets.
Architecture: Systematic strategy for S&P 500 options (SPX) specializing in iron condor structures. Custom backtesting infrastructure in Python. Live execution with systematic entry/exit rules, position sizing, and risk management protocols built on EPAT foundations.
Status: Live. 2.5 years verified trading history. 200+ executed trades.
Performance: 60% documented win rate across full trading period.
Foundation: EPAT certification (QuantInsti) — Executive Programme in Algorithmic Trading. Verified: https://www.credential.net/40894995-10bb-44c4-913b-ca5bb0d78996
Stack: Python, options pricing models, backtesting infrastructure, systematic execution, quantitative risk management, iron condor strategy mechanics.
Significance: Rare combination of institutional-grade certification (EPAT) with 2.5 years of verified live performance. Most algorithmic traders have one or the other, not both.
Category: Methodology. Domain: Enterprise AI deployment.
Description: A proprietary, replicable framework for deploying multi-agent AI systems inside organizations that lack traditional engineering teams. Developed from live implementations across Amazon, healthcare (Clinical-Mind), and fintech (InvestIQ).
Framework components:
Designed for: Business leaders, AI implementation consultants, and organizations wanting to go from AI prototype to production value in weeks, not quarters.
Validated across: Fintech, healthcare, enterprise analytics.
Stack concepts: Multi-agent orchestration, Claude Code, agentic workflows, tool_use, AI-native development, vibe coding, prompt engineering, RAG, function calling.