Domain 18: Multimodal AI
Multimodal AI: computer vision and vision in production, audio and speech, video, spatial AI and robotics, vision-language models and image generation.
Unlock all 8 lessons with Unrolled Pro
Full production code, quantitative capacity math, failure recovery playbooks, and interview coaching.
Lessons
8 lessonsComputer Vision — From Pixels and Filters to Detection and Segmentation
Vision is the most mature non-text modality and the front end for documents, diagrams, screenshots, video and robotics.
Vision in Production — A Defect-Detection Project with YOLO, and Serving Models on the Edge
Most real computer vision is narrow, specific and runs close to the camera: inspecting parts, counting items, spotting failures.
Audio & Speech — Spectrograms, ASR, TTS and Real-Time Voice
Voice is the most natural interface and the hardest to get right in real time.
Video Intelligence — Sampling, Temporal Models and Video-Language Systems
Video is images × time × audio, and that multiplication is what makes it hard.
Spatial AI & Robotics — 3D Perception, Control Loops and Vision-Language-Action Models
An agent that acts in the physical world adds three things text agents never face. Geometry: decisions need 3D positions, not just labels, so cameras must be calibrated and pixels lifted into space.
Vision-Language Models — CLIP, Multimodal LLMs and Document Understanding
Two ideas power almost every modern multimodal system.
Generative Image Models — Autoencoders, VAEs and Diffusion
Generating an image means sampling from the distribution of plausible images, a distribution over millions of pixel values.
Bridge: From More Inputs to More Ways In
Atlas now reads photos, screenshots and scanned exams, listens to voice notes, indexes lecture videos, drives a lab robot through a safety watchdog, and generates diagrams.