Ron Jailall - Resume

Professional Profile

Computer Vision Engineer & 3D Researcher with 15+ years of experience engineering high-performance ML solutions and interactive 3D graphics. Expert in VisionOS, Gaussian Splatting, and On-Device Inference, with a proven ability to optimize state-of-the-art algorithms for mobile Apple Silicon constraints. Deeply experienced in 3D geometry, shader programming (Metal/GLSL), and deep learning frameworks (PyTorch/TensorFlow). Passionate about pushing the boundaries of spatial computing and immersive technologies.

Core Competencies

3D Computer Vision: Gaussian Splatting (SHARP), Volumetric Reconstruction, View Synthesis, 3D Geometry.

Apple Ecosystem: VisionOS (RealityKit), CoreML, Metal Shaders, Swift, Apple Silicon Memory Management.

Machine Learning: TensorFlow 2, PyTorch, Model Conversion (ONNX to CoreML), Tensor Optimization.

Math & Physics: Linear Algebra, Projection Matrices, Shader-based Physics Simulation, DSP.

Featured Project: Interactive Volumetric Reconstruction on VisionOS

Real-Time Gaussian Splatting with Physics (SHARP Model Adaptation)

Algorithm Implementation: Adapted Apple's state-of-the-art SHARP model (Gaussian Splatting) to run natively on VisionOS, enabling volumetric 3D scene reconstruction from single 2D images.
Mobile Optimization: Overcame strict browser/WebGL memory sandboxing limits by re-architecting the pipeline for CoreML and native execution, ensuring reliable performance on the Apple Vision Pro.
Interactive Shaders: Designed custom vertex and fragment shaders to implement "jiggle physics," allowing users to physically interact with (tap, bounce) the reconstructed volumetric scenes in AR space.
Pipeline Engineering: Managed complex tensor conversions and input/output optimizations to successfully translate the research model into a production-ready CoreML package.

Professional Experience

ML Engineering Consultant / CV Researcher

2024 – Present

Remote

Specializing in 3D Computer Vision, efficient mobile inference, and Apple ecosystem development.

Efficient Human Matting (Matte Model Project): Implemented a state-of-the-art, CPU-friendly portrait matting system (MODNet/MobileNetV2) using TensorFlow 2/Keras, specifically architected for efficiency on consumer hardware. Engineered the end-to-end pipeline from dataset curation (P3M-10k) to ONNX export, enabling low-latency inference that rivals proprietary SDKs.
Procedural 3D Geometry (Midnight City Sailing): Developed a WebGL-based procedural generation engine to render a 3D neo-gothic cityscape, utilizing advanced shader techniques for atmospheric effects and geometry instancing. Applied rigorous linear algebra and 3D geometry principles to manage camera matrices, vertex transformations, and real-time rendering constraints.
Nvidia Jetson & Edge AI: Optimized computer vision models for embedded Nvidia Jetson platforms using TensorRT, achieving significant performance gains for real-time camera tracking systems. Created optimized Python/CUDA code for real-time video warping and blending.

Lead Engineer, AI R&D

2023 – 2024

Vidable.ai | Remote

Led research and implementation of Generative AI and ML infrastructure.

Research to Production: Collaborated with PhD researchers to evaluate emerging generative AI techniques and translate them into production-grade features. Modified and optimized C/C++ inference engines (llama.cpp, Stable Diffusion Turbo) to run efficiently on varied hardware targets.
Presentation & Communication: Hosted weekly cross-company technical seminars to explain complex AI inventions and research papers to engineering and product teams. Authored technical deep-dives on topics like On-Device VLMs, explaining the architecture of multimodal models for edge devices.

Lead Engineer

2014 – 2023

Sonic Foundry | Remote

Engineering leadership for large-scale video data pipelines and cloud infrastructure.

Video & Image Processing: Built and modified U-Nets for image analysis using PyTorch, applying deep learning to video segmentation tasks. Prototyped neural search of video archives using segmentation and classification algorithms (PyTorch, TensorFlow).
Hardware Control: Developed a Software Defined Radio (SDR) tool using DSP (Digital Signal Processing) to demodulate and decode RF control protocols, demonstrating strong signal analysis skills. Ported OpenGL ES 2 shaders to run on embedded development boards (BeagleBone Black).

Selected Talks & Publications

Apple's On-Device VLM: The Future of Multimodal AI

Conference Talk, Sep 2025

Presented a technical analysis of running Vision Language Models locally on Apple hardware, discussing memory management and quantization strategies.

Hyperfast AI: Rethinking Design for 1000 tokens/s

AI Tinkerers Raleigh, Dec 2025

Talk on the implications of ultra-low latency inference for real-time interactive systems.