R&D Log
Deep dives, tutorials, and research updates on Quantum ML, Physics-informed AI, and frontier engineering. Published openly.
An accessible deep-dive into VQE circuit design, ansatz selection, and ML-guided optimization landscapes for practitioners coming from classical machine learning.
A multi-round code generation pipeline that integrates Bandit static analysis directly into the LLM generation loop, using an adaptive security profile to dynamically re-weight prompt instructions based on detected vulnerabilities—driving output toward progressively safer code without fine-tuning.
Lessons from building an LLM-powered research assistant that processes 10,000+ papers. We cover RAG architecture, domain-specific embedding strategies, and the agent patterns that work in scientific contexts.
A comprehensive introduction to PINNs — neural networks that learn to satisfy physical laws. We cover the core theory, implementation from scratch, and real-world applications in fluid dynamics.
How I used IBM Qiskit and graphene quantum dot photonics to design a point-of-care biosensor capable of detecting neonatal sepsis biomarkers at 0.02 ng/mL — and what this means for diagnostics in low-resource settings.
An introduction to PSMT — a transformer architecture that dynamically modifies its own weights and topology during inference, achieving 37.4% better diagnostic generalization and 2.6x faster convergence on medical benchmarks.
A deep dive into designing HALO-UNet — a multistage deep learning framework for thyroid nodule segmentation that achieves 71% Dice score while running 70% faster than baseline UNet on resource-constrained hardware.