Demos

Transform a book chapter into a full multi-voice audiobook experience with this demo. Users upload a PDF, and the system uses an LLM to parse the text, map each character to a distinct voice with a matching personality, and reformat the prose into a dialogue-style script. Each line is then rendered using Gemini TTS, producing individual audio segments that are stitched together into a single narration, complete with a synchronized script, color-coded turn timeline, and a character cast map.

Demonstrated with a chapter from L. Frank Baum’s The Wonderful Wizard of Oz.

This demo is a “you might also like" recommendation engine built on a catalogue of toys.

Users can browse the catalogue and receive instant similar-item suggestions powered by TF-IDF content similarity, a technique that analyzes and compares product descriptions to surface the most relevant matches. This demo showcases how content-based filtering can be applied to e-commerce settings to improve product discovery and boost engagement without requiring user history or behavioral data.

Demo of a recruitment tool built to streamline hiring processes. Employers upload a job description in PDF, and JAME automatically extracts job title, company, and requirements, and generates a shareable application link where candidates can submit their CVs. Using large language models, JAME evaluates each applicant against the job requirements and produces a live ranked leaderboard with transparent, reasoned analysis highlighting each candidate's strengths, weaknesses, and overall fit, helping recruiters make faster, more informed hiring decisions.

A fun AI image generation demo where users can select or upload pet pictures, choose up to four clothing items, and have the AI generate a photorealistic image of the pet wearing those outfits in a charming, animal-appropriate way.

Under the hood, the selected clothing images are composited into a labeled grid and sent alongside the pet photo to a multimodal image generation model, which uses structured text prompts and behavioral guidelines to ensure anatomically correct, whimsical results. Beyond the novelty use case, the underlying technology has practical business applications such as virtual try-on for e-commerce and interactive visual marketing campaigns.

DocInsight converts a text-based PDF into structured, presentation-ready outputs: slides, a roadmap, flash cards, or a mind map. The workflow extracts readable text from the uploaded PDF, prompts the DeepSeek model to return strict JSON, validates that response against an expected schema, and renders the result with export options.

Its intended applications include training and enablement (turning internal guides and onboarding material into study aids), executive summarization of long reports and research papers, and knowledge organization for consultants, educators, analysts, and operations teams.

This demo is an AI-powered CV reformatting tool aimed at consultants working in international development.

Users upload their CV in PDF, DOCX, DOC, or TXT format, and an LLM extracts every detail - personal info, education, employment history, languages, and certifications - into a structured profile, which is then injected into official CV templates according to EU, UN, and World Bank formats. All required fonts, borders, spacing, and table layouts are preserved exactly as mandated. The result is a one-click solution to a tedious but high-stakes formatting task that consultants frequently face when bidding on international contracts.

A humorous AI image-understanding demo where users upload a photo of their food and receive a brutally honest critique from "Tante Fawzia," an Egyptian grandmother persona with fifty years of cooking experience and zero tolerance for mediocre plating.

A vision model analyzes the actual visual content of the dish - presentation, colors, plating, and portions - while Tante Fawzia's personality, including her Egyptian Arabic interjections and habit of comparing everything to her own cooking, is crafted entirely through prompt engineering.

Beyond the entertainment value, the underlying tech demonstrates practical business applications like automated image categorization and quality control, where visual AI can assess products against brand standards and flag items that fall short.

Second Opinion show cases "Edge AI", running a large language model entirely inside your web browser via WebGPU hardware acceleration, so it can analyze product reviews and generate a private summary without ever sending data to a server.

Technically, it works by checking for WebGPU support and using Apache TVM and WebLLM to compile the model for your local graphics hardware, downloading and caching the model weights in the browser's IndexedDB for instant loading on return visits, then scraping the on-page review text and streaming locally-computed tokens back to the UI in real time. Its two main selling points are absolute data privacy (sensitive data like HR reviews or medical notes never leaves the device) and zero cloud computing costs, since the workload runs on the user's own GPU rather than incurring cloud API fees.

A satirical AI chatbot demo styled as a luxury gym concierge catering to "deeply unserious fitness people." Users can ask about classes, coaches, packages, policies, or the most refined way to avoid measurable exertion, with the assistant delivering grounded, context-aware answers through an API-first RAG (Retrieval-Augmented Generation) system backed by a curated knowledge base and a FastAPI backend.

Beyond the humor, the demo illustrates real-world business applications of RAG, including customer support automation, sales enablement. This is especially interesting for document-heavy industries like legal, healthcare, and finance, where AI can provide accurate answers grounded in trusted internal documents while keeping data confidential.

This demo is a localized AI grocery assistant tailored for the Egyptian market. It acts as a smart shopping companion that helps users discover recipes, access nutrition facts, build meal plans, and manage their grocery cart.

It is built on A2A Protocol v1.0, the open standard for agent-to-agent communication. A series of specialized agents communicating together to collaborate in real time to search products, manage your cart, plan meals and list meal nutrition facts.

This demo illustrates how AI can be embedded into everyday retail and grocery experiences to make shopping a guided, personalized interaction with clear potential applications for supermarkets, food delivery platforms, and e-commerce grocery stores in the region.

Rawy is a personalized daily audio news platform that transforms global headlines into a podcast tailored to each user's interests, using AI narration and verified sources.

Rather than endlessly scrolling through news feeds, users receive a curated daily briefing delivered in a clear, warm, story-first format, currently available via a beta program. Rawy combines the personalization of algorithmic news curation with the convenience and immersiveness of a podcast listening experience.