Multi-Agent System
This prototype for a fictitious company, Cactus Financial, demonstrates how a team of AI agents can automate a common business challenge: managing customer support at scale. Functioning less like individual bots and more like a coordinated department, the system uses Model Context Protocols (MCPs) to:
- Triage and prioritize new tickets by category and risk.
- Generate brand-aligned responses by securely accessing an internal knowledge base.
- Escalate complex or sensitive issues to human agents when required.
The system is designed with a full suite of operational features, including an AI co-pilot for human agents, an analytics dashboard for KPIs, and comprehensive error monitoring for reliability.
Explore the prototype here. You can seed a ticket and observe its journey through the system, from the initial queue to the human escalation dashboard.
Chef Companion
The Ayurvedic Chef Companion is a trustworthy conversational agent that creates simple, authentic recipes based on user-provided ingredients.
The core technical challenge was to ensure all information conforms strictly to Ayurvedic principles and is 100% accurate, safe, and free from AI "hallucination."
Key Capabilities:
- Verified & Trusted Knowledge: The AI's knowledge is exclusively sourced from a private, curated library of content provided by an Ayurvedic professional. This use of a vector store guarantees that every recipe and piece of advice is authentic and reliable.
- Dynamic Recipe Generation: The agent intelligently assesses a user's ingredients, identifies missing components for a balanced meal, and generates a complete recipe with shopping lists and step-by-step instructions.
- Principled & Safe by Design: The system is built with a deep set of constraints, preventing it from ever suggesting non-Ayurvedic ingredients and ensuring it adheres to core cooking principles.
- Educational by Design: The agent explains the "why" behind key concepts, equipping users with practical knowledge to build sustainable, healthy eating practices.
News Aggregator
Every morning, we're hit with a flood of information. How do you find the signal in the noise? I designed this autonomous AI agent to solve that problem by distilling the day's essential news into a single, insightful intelligence briefing.
Key Capabilities:
- Autonomous Daily Briefing: The agent runs automatically every morning, scanning multiple news APIs and delivering a fresh, organized summary to a custom Notion page without any human intervention.
- Cross-Category Insight Generation: More than just a summarizer, the agent performs a high-level analysis to find patterns, emerging trends, and hidden connections between seemingly unrelated topics.
- Structured & Actionable Output: The briefing is delivered in a clean, user-friendly template with collapsible sections, allowing for a quick overview or a deeper dive into specific areas.
- Source Attribution & Transparency: To ensure credibility, the agent meticulously cites all original sources at the bottom of each report, allowing for easy verification.
- Multi-API Integration: The agent seamlessly connects to and processes data from multiple external news sources, ensuring a comprehensive and diverse range of information.
You can see an example of the daily briefing the agent produces via this link.
