About Me
I am Latchan, and my work centers on architecting hardware-efficient, scalable deep learning systems for high-degradation and extreme physical environments. I specialize in bridging continuous physical dynamics with discrete neural computation, utilizing sub-quadratic mechanisms (SSMs), uncertainty-aware learning, and geometric residual frameworks. Driven by end-to-end ML workflows, I actively open-source my architectures and focus on translating theoretical foundations into provably reliable, real-world physical deployments.
Core Research Themes
- State Space Models & Sequence Modeling: Optimizing 2D and 3D Mamba architectures for spatial, temporal, and volumetric modeling.
- Uncertainty Quantification: Integrating conformal prediction and fuzzy logic for provable reliability in medical and satellite imagery.
- Efficient Architectures: Developing lightweight, high-throughput architectures via geometric residual learning and DDL overwrites for edge-device deployment.
- Medical Imaging & Remote Sensing: Developing robust frameworks resistant to distribution shifts and data leakage. Utilizing light-weight & efficient architectures for remote sensing and earth observations.
- Computer Vision & OCR: Generative Adversarial Networks, Diffusion Models, Tesseract, Vision Transformers, and Deep Delta Learning applications for image restorations, degraded document analysis & scene text recognition.
Independent Research Group — Halo Mind
Hybrid Architectures & Lightweight Optimization | Machine Intelligence & Neural Dynamics
I am the Co-Founder and Chief AI Scientist of Halo Mind, a Research Group & applied DeepTech and AI research firm focused on hardware-efficient systems, AI systems, State Space Models (Mamba), Deep Delta Learning (DDL), uncertainty-aware machine learning, and lightweight architectures for real-world deployment.
Our research spans medical imaging, remote sensing, space-weather forecasting, OCR, and robust machine learning under resource-constrained environments. We emphasize rigorous experimentation, reproducibility, and mathematically grounded AI systems capable of operating under distribution shifts and uncertain conditions.
Research Areas: State Space Models (Mamba), Deep Delta Learning, Conformal Prediction, Computer Vision, Medical AI, Remote Sensing, OCR, Edge AI, and much more…
Website: Halo Mind Research Group