Work

I enjoy solving problems and don't shy away from tasks that initially seem beyond my reach. That said, I'm a strong believer in collaboration — sometimes you need another set of eyes to unlock the solution or see what you've been missing.

In this era of AI-powered development, my workflow has evolved significantly. I start by brainstorming with AI tools like Claude or Grok, working through ideas and exploring different approaches. Once I have a clear direction, I develop a concrete plan of action. From there, documentation and organization become critical — I rely heavily on Notion to keep track of documents, plans, and specific details that might otherwise get lost in the development process.

For ambitious tasks like refactoring or heavy backend work, I use Claude Code with multiple agents tackling different context-specific areas. For frontend development, I lean on Cursor. As someone with minimal frontend experience, I've found it to be an incredible learning tool that helps me create polished interfaces that complement the backend work.

My philosophy on AI-powered development: In 2025 (almost 2026), you need to leverage AI tools to their fullest potential — but you need to use them as the tools they are. Minimal effort produces minimal results (what the industry calls "slop"). Instead, use AI intentionally to learn, enhance your skills, and develop a strong sense of product design and quality. The tools are powerful, but your taste, judgment, and effort are what turn that power into something meaningful.

Data Analyst / Software Engineer with 2+ years of experience spanning data science, machine learning, and full-stack development.

I started my career in data science and analytics at a blockchain analytics startup, where I focused on anomaly detection patterns for predicting fraud and threats. Beyond the technical work, I served in a client-facing role, communicating data insights to organizations including the Department of Defense, national banks, crypto exchanges, and securities commissions. This experience taught me how to translate complex technical findings into actionable intelligence for diverse stakeholders.

I then moved into a software company developing NLP SMS agents for verticals like insurance and healthcare. There, I worked hands-on with machine learning models, building automated Python evaluation frameworks to analyze model performance in both test and production environments. Through A/B testing and detailed confusion matrix analysis, I identified where our models were mishandling data and contributed to improving F1 scores by upwards of 20%. Working on the Machine Learning team gave me the opportunity to collaborate closely with engineering and production departments, bridging the gap between model development and real-world deployment.

Currently, I'm building Atticus AI, a legal AI platform powered by Claude. The mission is to provide SMB legal professionals and lawyers with a way to leverage AI effectively and safely, without worrying about confidentiality breaches or compliance issues. I've handled the full-stack development of Atticus AI, shipping features including document creation and iteration, RAG context integration (pulling directly from users' documents) into the AI assistant, document analysis, and real-time search. Atticus AI is actively in development, and I'm excited to share more as it evolves.