I am a passionate engineer and researcher driven by the challenge of solving complex, open-ended problems at the intersection of software and machine learning. With a strong foundation in algorithms and systems from my time as a competitive programmer, I now focus on building and analyzing intelligent systems. My recent work involves exploring the nuances of human-AI interaction, particularly understanding the gap between user expectations and the capabilities of large language models.
My most recent research project, "The Conversational AI Tax," is a deep dive into the unique failure modes of mental health chatbots.
- 🧠 Analyzed 30,000+ user reviews to build a data-driven "taxonomy of failure."
- 🤖 Leveraged BERTopic and RoBERTa to move beyond keywords and understand the semantic meaning and emotional intensity of user complaints.
- 📊 Proved a key insight: Failures in AI performance (like memory and personality) are significantly more emotionally damaging to users than high prices.
- 📈 Discovered a "smoking gun" linking a real-world product update to a massive, 300% spike in specific AI-related complaints.
- Research & Analysis: Completing my Certificate in Computer and Data Science at MIT's Emerging Talent Program, where I'm how to Interact with real-world datasets and engineer an Impactfull solutions.
- Core Engineering: Continuously honing my skills in algorithms and data structures through competitive programming on Leetcode and Codeforces.
- Exploring: Investigating the latest developments in model fine-tuning, retrieval-augmented generation (RAG), and AI safety.
- Languages: Python, JavaScript/TypeScript, Java, C++
- AI / ML: PyTorch, Scikit-learn, Transformers, LangChain, BERTopic
- Web & Backend: Node.js, React.js, Express.js, SQL/NoSQL
- Tools: Git, Docker, Google Cloud, Jupyter Notebooks
