AI Developer • Solutions Engineer • Reasoning Engine Architect

"Building Reasoning Engines that move beyond simple scripts."

I am an AI Developer and former Solutions Engineer focused on architecting production-grade ML systems and autonomous agent frameworks. I specialize in building Python-driven agents that plan and execute complex tasks with strict validation layers to eliminate "hallucination" and ensure data integrity.

An orchestration layer for autonomous financial intelligence.

FinSurf is a decentralized network of specialized AI agents designed to navigate the complexities of financial data. It functions as a high-integrity research team, gathering and validating market data in real-time.

• The Goal: Transform raw financial data into professional-grade reports using a multi-agent network.
• The Architecture: A Directed Acyclic Graph (DAG) orchestrated via LangGraph, featuring parallel execution and conditional routing (e.g., skipping dividend analysis for non-dividend stocks to save costs).
• Security Layer: Implemented Semnatic Guardrails that use a dedicated Security Agent to sanitize inputs and block prompt injections before any high-cost API calls are made.
• Technical Edge: Implemented strict Python validation layers and automated RAG (Retrieval-Augmented Generation) to ensure high-fidelity outputs.
• Tech Stack: Python, Pandas, LangChain, React 19, Tailwind CSS 4, Express.

• 🔭 Current Project: FinSurf - Expanding the Profit Navigator for historical price retrieval and projected exit strategies.
• 🌱 Learning Path: Deep-diving into ML/LLM fundamentals, Cloud Infrastructure (AWS/Docker), and Higher Mathematics.
• 🛡️ Security Focus: Developing advanced "Prompt Shield" architectures to protect internal agent instructions from being leaked via indirect injections.

Building FinSurf was an exercise in bridging cutting-edge AI orchestration with professional-grade document engineering.

• Isolation of Concerns: By separating Research, Tax, Dividend, and Sentiment logic, we reduced "prompt drift" and improved output accuracy.
• Strict Validation: Implementing mathematical precision for the Dividend Agent taught the importance of "Python-first" validation layers—calculating fractional shares and cumulative totals programmatically rather than relying on LLM arithmetic.

• Color Conversion: Since libraries like html2canvas do not natively support oklch or CSS variables, I developed a custom color-conversion utility to ensure high-fidelity rendering.
• Adaptive Pagination: Developed logic to dynamically switch between "Single Page if Fits" and "Multi-page" layouts, optimizing the reading experience for digital reports.

• Containerization: Integrated Docker to ensure environment parity across local development and cloud deployment.
• Telemetry: Built a SQLite-backed cost-tracking system to monitor per-agent token usage and enforce daily budget caps.

I’m looking to partner with other builders focused on automating repetitive and mentally taxing tasks.

• Ask Me Anything: I'm happy to learn from your expertise or help debug your agentic workflows!
• Reach Out: [email protected]
• Fun Fact: I love pandas (the library and the animal) and enjoy racing down dirt roads at a blistering 1 m/s. 🤠