Featured Work
I solve complex problems by asking the right questions, thinking creatively about AI systems, and ensuring technology serves people—not the other way around.
Human Preference Modeling Framework
The Challenge
Current AI systems optimize for metrics that don't capture what humans actually want. A "helpful" response varies dramatically based on context, expertise level, cultural background, and individual preferences.
Through data annotation work, I discovered that expert disagreement isn't noise—it's the most valuable signal for understanding human preference diversity.
The Solution
A multi-dimensional preference modeling system that captures context-dependent human values instead of flattening them into binary good/bad ratings.
Key Innovation
Instead of asking "Is this response good?", we ask "For whom, in what context, and according to which values?"
Context Modeling
Capturing situational factors that influence preference: urgency, expertise level, cultural context
Value Alignment
Identifying underlying human values that drive preferences: efficiency vs. thoroughness, creativity vs. accuracy
Dynamic Adaptation
AI systems that learn and adapt to individual user preferences without losing generalizability
All Projects
The Subjectivity of AI Quality
How data annotation work revealed that expert disagreement is a feature, not a bug—and what this means for AI training.
AI Model Training Lifecycle
Watch me explain the complex AI model training process using intuitive cooking analogies that make technical concepts accessible.
Human-Centered AI Strategy
A strategic framework for building AI products that prioritize human values alongside technical capabilities.
SignalStack Website
The source code for this website built with Next.js, showcasing modern web development and AI-forward design patterns.
AI Insights & Analysis
Deep dives into AI development, human preference modeling, and building better AI systems through thoughtful analysis.
Let's Collaborate
Interested in discussing AI, human-centered design, or potential collaborations? I'd love to connect.
See My Work in Action
AI Model Training Lifecycle
Watch me break down the complex AI model training process using an intuitive cooking analogy. This demo explains everything from data preparation to deployment in accessible terms that anyone can understand.
Key Highlights: Pre-training data strategy, distributed model training, post-training optimization, safety evaluation, and deployment coordination—all explained through cooking metaphors.
AI Model Training Lifecycle • Educational Demo
This demo showcases my ability to explain complex AI concepts in accessible ways, making technical processes understandable for diverse audiences.
Watch on YouTube →Interested in Collaborating?
I'm always open to discussing these projects, sharing insights, or exploring potential collaborations on human-centered AI research.