Browse applications built on modern technologies. Explore PoC and MVP applications created by our community and discover innovative use cases for modern technologies.
pre-purchase fit-assurance shop. The question it kills is "will this actually fit my room?" — the single biggest reason furniture bought online comes back.
LineGuard spots machines drifting toward failure about a day earlier than fixed alarms, predicts when they will trip, and an AI copilot on AMD GPUs explains why, cites the plant's manuals and drafts a work order: breakdowns become planned repairs.
InnerOS turns AMD-powered AI into governed real-world action. Gemma reasons, Evolus runs workflows, and Alexa/Inner Oz can query live operations and create governed GitHub tasks for AI agents, with approvals, evidence, and replay.
Invictus Earth2 Keepers is a three opencode AI agent that seats on GLM-5.2 (Featherless) plan, build and 170 numbered requirements verified, two fix rounds after review, about 2 h 50 min with no human steering, for $13.83 of model spend.
The Roast Auditor is a voice accountability agent: tell it what you actually did today, and it cross-references your own words against your commitments, calls out vagueness and contradictions live, and gives you a blunt final verdict.
A governed realtime voice agent powered by AssemblyAI that reads real systems, reasons locally, requests human approval, executes actions once, verifies the result, and keeps auditable proof.
Autonomous zero-downtime database migration rehearsal, query blast-radius analysis, and compatibility view healing engine built with IBM Bob 2.0.
DevForge is a multi-agent pipeline that plans, builds, tests, debugs, and security-reviews code changes on its own, then pauses for a human to approve before release , handing back a real, downloadable deliverable plus a full audit trail.
SmartCare AI uses an IBM Granite 3.0 multi-agent pipeline and hardware QR codes converted to url to diagnose appliance fault codes instantly, classify electrical hazards, and provide safe DIY repair playbooks or automated technician escalation.
ShowMe is the AGI , it teaches any agent a workflow a human can show. Demonstrate it twice. ShowMe extracts what changes, compiles a skill, and IBM Bob, Cursor, Claude, or Codex repeat that job exactly.
BugForge is an AI-driven chaos engineering platform that uses LLMs to generate compound fault-injection scenarios, executes them against live Docker containers, and produces structured resilience reports with real-time metrics and pass/fail analysis.
Describe an automation in plain language. Get a real, executable, visually editable workflow whose agent nodes reason and decide while it runs.
A centralized GitHub Actions workflow scheduler and on-demand dispatch management platform that replaces unreliable native cron triggers with UTC scheduling, automated organization-wide workflow discovery, token-bucket rate limiting, and an audit trail.
AI-powered developer workflow automation that detects bugs, analyzes root causes, applies fixes, runs real tests, verifies changes, and generates actionable reports.
SpeakEasy is a multilingual real-time AI voice assistant that understands natural speech and uses tool calling to control fan and light functions through a browser-based interface.
Intel Bimanual VLA Manipulation with Multi-Modal Reasoning Track
Two simulated SO-101 arms in MuJoCo set a dinner table from a spoken-style command: an OpenVINO INT4 LLM plans, an INT8 camera U-Net localises objects, and IK skills with a staged hand-off execute. 10/10 randomized seeds.
Two SO-101 arms in MuJoCo turn one spoken-style command into a full table setup: drawer, fork hand-over and a two-arm pour. A learned ACT policy runs on OpenVINO INT8 at 1.35 ms on an Intel CPU, 5.2x faster than PyTorch.
Two SO-101 arms relay a cup across a dinner table in MuJoCo from a language instruction. SmolVLA fine-tune, OpenVINO FP16/INT8 export, closed-loop eval on a bare-metal Core Ultra X7 358H + Arc B390 — with the failures measured, not hidden.
RecoverVLA turns a table-setting instruction into coordinated dual SO-101 skills in MuJoCo, retries failed contacts, and reports seeded success instead of a one-shot demo.
Two simulated SO-101 arms set a dinner place from a spoken-style instruction with a fine-tuned SmolVLA policy running on an Intel iGPU through OpenVINO, while a physics-aware supervisor catches drops and retries safely.
Two SO-101 arms that set your table and pour your drink on a spoken request, and stop mid-pour on one word. For people who can speak but can't reach. Voice by Speechmatics in real time; planning and a separate safety check Intel OpenVINO with Qwen2-VL.
HomeAngel AI is a privacy-first edge fall-detection system that analyzes video locally on SiMa.ai Modalix and sends only text JSON/Telegram alerts to family.
TableTogether uses a trained vision-language policy to coordinate two simulated SO-101 arms for table setting in MuJoCo, with local OpenVINO inference, a live interface, and reproducible evaluation.
Two SO-101 arms setting a dinner table in MuJoCo. Every skill is gated on physical postconditions, not script completion: force-bearing contact, sustained lift, stable release. Two arms lift a plate that one arm can only lever onto its rim.
CoAct is a bimanual robotics system where two embodied agents independently interpret spoken instructions, detect disagreement before acting, and use real-time Speechmatics voice correction to safely resolve ambiguity before physical commitment.
Two SO-101 arms set a dinner table from a spoken command. A 1.5B model parses intent locally in OpenVINO INT4; a deterministic gate checks every step against the measured scene and refuses what it cannot ground. Scored on 20 seeds, failures published.
A learned bimanual manipulation system where one SmolVLA policy controls two SO-101 arms from three RGB views, language instructions and a 12-D robot state to complete plate, fork and mug placement in MuJoCo.
Blinky is an AssemblyAI voice-controlled desktop tutor that sees your screen, automates tasks, edits videos, searches live web, plays media, summarizes WhatsApp, and works from mobile.
EdgeWatch monitors simulated IoT fleets, analyzes telemetry with AI, and predicts device failures before they happen through a lightweight edge-to-AI pipeline.