Project directory

Academic timetabling tool where an LLM turns messy files and requests into constraints and CP-SAT or a C++ solver builds verified schedules.

Checks packing photos for Indonesian online sellers and flags weak dispute evidence, with retake advice and buyer-claim triage.

Lets clients or their AI agents commission field professionals to photograph tagged items, with AI tag reading and human review for audit reports.

Web IDE that shows coding agents' edits on a live repository graph with impact cones and requires human approval before merging to main.

Phone web app that samples camera frames, detects path obstacles with Meta vision models, and speaks navigation cues to blind users.

A phone game where an AI agent turns photos taken on a walk into a mystery case, solo or with a friend, solved by exploring on foot.

Coordinator that turns consented voice and photo field reports in Bangla, English and Arabic into evidence-linked drafts for human review.

Mobile cozy game teaching clinical reasoning and health literacy through persistent fictional patients who remember the player's decisions.

Agent that answers questions over Kosovo public procurement data and logs every action in a hash-chained, verifiable audit receipt register.

Turns one phone photo into a WooCommerce draft listing, studio photos and Instagram post in English and Bangla for small Bangladeshi sellers.

OpenEnv reinforcement learning environment where LLM agents classify and prioritize emails, including fake-urgency and phishing detection.

Builds structured courses from the most relevant YouTube clip spans for any topic, with verified recall questions, teach-back and a coach.

Smart-glasses app that reads a wine label and predicts by voice whether you and your companions will like it, based on your own taste profile.

Audits image annotations against independent Mask2Former segmentation, giving tri-state verdicts with uncertainty and SHA-256 receipts for CV teams.

Browser kernel that loads LLM-generated pure-function modules after differential verification, WASM sandboxing, signing and caching.

Simulates two side-by-side futures for a major life decision, with labeled facts vs assumptions, shocks, and chat with each future self.

Windows desktop agent that runs terminal commands, manages files, clicks GUI elements and sends email or WhatsApp with approval gates.

Turns voice notes, images and brain dumps into a single next action shown across watch, phone, web and WhatsApp for adults with ADHD.

Photo and voice meal logger that identifies Southeast Asian dishes and shows cited kidney-relevant nutrients for CKD patients and dietitians.

Proposed observability and self-healing layer for autonomous AI agents in production; the repo holds only a PDF and a short README.

Pipeline that segments images of electronics or industrial parts, diagnoses defects, and generates remediation scripts or G-code.

Rust CLI that lets coding agents express refactors as compact semantic patches, lowered to Git diffs and applied, plus reverse compression.

Generates illustrated multi-chapter stories with figures from a prompt, with an offline local-model fallback for low-connectivity users.

Appears to be a hub for connecting multiple AI models such as Muse, Claude and GPT; the repository contains only a title README.