Explorations in aerospace engineering, AI/ML, computational fluid dynamics, and the future of technology.
An AI model wrote the aircraft geometry, the fluid solver, the GPU particle system, and the rendering pipeline — then packaged it all into a lightweight ~2.2 MB standalone HTML file that runs in any browser at 60 FPS. No CAD, no install, no plugin. Two procedurally-generated aircraft (a refined airliner and a stealth UCAV), controllable GPU tracers, smoke visualization, and emergent wingtip vortices — all generated by Kimi K3 from a natural-language prompt.
Read Article →
Six sparse-regression frameworks, automatic multi-object tracking, fluid-physics outputs, and bundled PyTorch — plus AeroFlow NS, a real-time 3D Navier-Stokes wind tunnel in your browser with GPU tracers, smoke visualization, and a stealth UCAV model.
Read →
Modern design pipelines optimize external shapes and structures — often with simulations trained on simulations — while the systems integration, certification, manufacturability, and envelope-corner knowledge that turns a design into a program sits below the waterline.
Read →
A candid post-mortem of an LLM-run smoke-tunnel validation — the confidently-wrong drag number, the 0 m/s artifact, the failed vision detection — and why setup, execution, and interpretation still demand careful human hands.
Read →
A fluids engineer directing an LLM through natural-language conversation produced a complete OpenFOAM simulation, parametric shape optimization, and automated ParaView visualization — on a bear geometry.
Read →
Physics-informed neural networks trained on 1959 wind tunnel data capture missing physics that traditional CFD simulations miss entirely — on a single-board computer.
Read →
How AI-powered development enabled the creation of SkyView Pro, a comprehensive real-time geospatial platform using only open-source data.
Read →
Smoke flow visualization, optical flow velocity vectors, and SINDy analysis on a 1:64 scale car model — all from a small wind tunnel at home.
Read →
Why specialized tools developed by engineering firms over decades often outperform modern flashy alternatives that deviate from real-world physics.
Read →
A tool for precise analysis of movement in video using Dlib — for loading videos, calibrating measurements, selecting regions of interest, and deriving velocity and displacement data.
Read →
Integrating advanced computer vision (Optical Flow) with data-driven modeling (SINDy) to understand and predict fluid dynamics from video data.
Read →
How a tiny energy-efficient powerhouse became the brain of a home automation system with Home Assistant, AI vision, and media serving.
Read →