Applied AI builder focused on workflow automation, agent memory, multimodal systems, and practical AI products.
I keep a deliberate build-and-study loop around the fast-moving AI stack: coding agents, MCP/tool use, retrieval and memory, evals, browser automation, multimodal models, and deployment patterns. The goal is not to chase every new tool. The goal is to test what is useful, understand the constraints, and turn it into working systems.
Portfolio: joshfriedman-dev.com
- Agent Memory Labs - early-production memory infrastructure for AI agents: persistent, searchable context across code, git history, web research, and conversations, exposed through CLI/API/dashboard/MCP surfaces.
- mutdashboard.com - unofficial Madden Ultimate Team auction house companion built around reverse-engineered data flows and practical user workflows.
- voice-computer-use-agent - local computer-use control for AI agents, optionally driven by voice input.
- Dictate - speech-to-text workflow tool with translation and customizable formatting.
- Radiology AI Video - multimodal workflow that turns patient radiology reports into patient-friendly avatar video explanations with review gating.
- 3D Render Radiology - MRI/CT scan to patient-specific 3D model workflow built around TotalSegmentator.
- web-capture-extension - Chrome extension for capturing webpages, documentation, screenshots, and structured context for downstream AI/LLM use.
- codex-mobile-app - mobile companion for connecting to coding agents and continuing agent conversations from a phone.
- VoiceClone TTS - Twitch extension concept for paid viewer-triggered AI voice/TTS playback during live streams.
- Evolution Engine - altered evolutionary timeline simulation plus AI-generated educational video pipeline.
- AI agents and tool-use workflows
- MCP, retrieval, memory, and GraphRAG
- Browser automation and computer-use agents
- Multimodal pipelines for text, image, audio, and video
- Practical product workflows with state, review gates, and delivery