PROJECT SCOPE
– Implement a lightweight language model (1-1.5B parameters) running locally on web and mobile browsers via WebAssembly (WASM).
– Enable the assistant to interpret and rewrite user inputs before sending them to a cloud-based model.
– Create a peer-to-peer VPN-like feature within the PWA for bypassing content blocks using WebRTC and STUN servers.
– Develop the PWA front-end using Vue.js
EXPECTED OUTCOMES
– Functional voice-based assistant prototype embedded in a PWA.
– High-quality responses, comparable to the original model.
– Local language model processing inputs on the browser and routing to a larger backend model.
– Peer-to-peer VPN-like functionality within the PWA, allowing users to navigate content blocks by choosing endpoints.
– Ability for users to route through one or multiple peers, similar to TOR.
– Seamless communication between local and cloud models.
– Real-time WebRTC connections between peers for data transfer.
– Low-latency voice interaction combining local and cloud processing.
– Seamless user experiences:
Use case 1: When a user clicks a button on a generic PWA on the browser, it will launch the assistant within the same page. The user can use text & voice as input and receive text & voice response in real-time from the assistant Maximum acceptable latency of the prototype is 6 seconds, with a feasible proposed solution to improve latency to under 3s for the next phase.
Use case 2: A user needs to chose from what country to connect as exit point , since an X amount of exit points per country will exist depending on the amount of peers live in that country at that moment. User can chose to A) NOT connect via peer-VPN B) exit via one exit point C) exit via 2 exit points ( slower but even more privacy) . If a peer goes down, or is overloaded it will automatically switch to the next one in that country.
REQUIREMENTS
– Experience in P2P PWA development.
– Proficiency with WebAssembly and running machine learning models on browsers.
– Familiarity with WebRTC and STUN server setup.
– Strong problem-solving skills and ability to deliver within tight timelines.
– Skilled in Vue.js development.
DEADLINE
Completion and demo by 26th July.
Budget: $2,000
Posted On: July 16, 2024 07:17 UTC
Category: Full Stack Development
Skills:Vue.js, AI App Development, AI Bot, AI Model Integration, WebAssembly, WebRTC
Country: Malta
click to apply
Powered by WPeMatico
