Deep learning pipeline for document layout understanding and OCR on historical texts.
I build ML systems, data pipelines, and research tooling.
I translate research ideas into scalable software with deep learning, multimodal modeling, and distributed systems.
Graduate researcher. Former co-founder at Neotix.
- Multimodal modeling and representation learning across vision and language.
- Deep learning pipelines for data curation, training, and evaluation.
- Distributed systems for large-scale datasets, versioning, and reproducibility.
- Built multimodal data pipelines for large-scale ML experimentation.
- Shipped research tooling for dataset curation, labeling, and analysis.
- Versioned datasets and training artifacts with scalable storage.
How I think about ML research, data, and infrastructure.
Three system views that reflect how I scope research problems, build data pipelines, and ship models.
Full project timeline
Recent ML research highlights are above. This is the complete catalog across research, systems, and software.
RenAIssance Document Analysis
End-to-end deep learning pipeline for layout understanding, OCR, and structured digitization of historical text.
ISSR — Crisis Detection
Social signal analysis for mental health crisis detection with sentiment modeling and geospatial insights.
MinIO + LakeFS Data Infrastructure
Scalable storage, dataset versioning, and lineage tracking for large ML experiments and artifacts.
ArtExtract — Art Analysis AI
CNN-RNN model for artwork classification, style detection, and similarity search across art history.
Additional projects (embodied AI, full-stack, mobile, games)
2025 - ML Research & Systems
Developed remote teleoperation system for bimanual dexterous manipulation and data collection. Implemented real-time control protocols for imitation learning tasks.
Engineered comprehensive data extraction pipeline implementing SLAM-ORB3 for trajectory extraction and real-time 7-DOF trajectory analysis.
Developed annotation tools for robotic manipulation datasets using embodied Chain-of-Thought methods for labeling gripper poses and keyframes.
Architected scalable data storage and versioning system for robotics datasets with Git-like version control for large sensor data.
Built end-to-end robotic learning platform with data collection pipelines and policy training infrastructure for rapid prototyping.
Google Summer of Code 2025 project. Built AI pipeline for digitizing Renaissance-era texts using deep learning for layout recognition and OCR.
GSoC candidate project. AI-powered system for suicide prevention through social media analysis, sentiment detection, and geospatial crisis mapping.
Deep learning system combining CNN-RNN architectures for artwork classification, style detection, and similarity analysis across art history.
2024 - Full-Stack Development
AI-powered health and nutrition recommendation system. Built full-stack application with intelligent meal planning and dietary analysis.
Flutter-based student application for discovering campus events, dining options, and university resources. Designed to enhance student life experience.
Comprehensive Java-based POS system with inventory management, multiple payment methods, and sales reporting. Built with modular architecture for retail environments.
2023 - Game Development
Modified version of the classic arcade game with enhanced gameplay mechanics and modern graphics. Implemented in C++ with custom game engine features.
Research notes and systems breakdowns
Short technical writeups on robotics, ML systems, and multimodal research.
The State of Robotics in 2025: Why the Hype Isn't Lying (Yet)
A deep dive into the four critical bottlenecks slowing the robotics revolution and why general-purpose robots remain inevitable
Making Sense of Multimodal Models with Partial Information Decomposition
A deep dive into how Partial Information Decomposition (PID) reveals how different modalities interact in AI systems, from redundancy to synergy