טופס פרויקט

MyPrivateAI

15006902
:מספר הפרויקט
ארז פוליאק
:שמות הסטודנטים המציגים
גב' קמחי-רוכברגר גילה
:שם המנחה
סדנת אפליקציות של Generative AI
:שם הסדנה
מסלול אפליקטיבי
:מסלול הסדנה
:GitHub
פוסטר
מצגת
:תקציר הפרויקט
MyPrivateAI is a local-first desktop application that lets users analyze sensitive documents with powerful AI without sending data to the cloud. It installs in one click, runs entirely offline, and requires no technical expertise.
Many professionals cannot use cloud AI tools because their documents contain confidential information. Existing local alternatives are either difficult to use or struggle with complex document reasoning — and local models typically perform far below frontier cloud models on demanding tasks.
Using a pipeline combining advanced RAG techniques and multi-turn reasoning agents, MyPrivateAI narrows this gap, enabling advanced multi-document analysis on consumer hardware. On the SciRAG-QA benchmark — which requires reasoning over external scientific documents rather than a model's built-in knowledge — it achieved 95% of GPT-5.1's performance while remaining fully private and offline.
By combining privacy, accessibility, and state-of-the-art document reasoning, MyPrivateAI makes powerful AI accessible to anyone working with confidential information.