Technical Architecture

JobScan is built upon a modular and scalable architecture designed to serve millions of users across diverse use cases while maintaining agility, performance, and security. The platform architecture is divided into five major components: the user interaction layer, AI semantic engine, microtask orchestration system, recruiter dashboard, and blockchain credentialing layer. Each component interacts through APIs and event-driven pipelines to deliver a unified, real-time experience.

The user interaction layer consists of the web platform (developed with React, TailwindCSS, and Next.js) and the Telegram bot interface (built with Node.js, Telegraf, and Firebase integration). All user input—resumes, skill declarations, job preferences—is routed through a secure gateway into JobScan’s semantic engine. This AI engine leverages large language models (LLMs), currently integrated with DeepSeek and OpenAI APIs, and includes JobScan’s proprietary vector embedding module. It encodes both candidate profiles and job postings into a shared latent space, enabling high-precision matching through similarity scoring and contextual keyword mapping.

The microtask orchestration system handles short-form paid tasks and manages both supply (incoming task listings) and demand (eligible users). Task complexity, requirements, and verification rules are encoded into templates that dynamically assign jobs to qualified users based on skill vectors, completion history, reputation, and availability. The scheduler balances real-time demand with latency guarantees, ensuring tasks are not bottlenecked and users are consistently engaged. A feedback loop allows user ratings to influence future task allocations and leveling.

On the recruiter side, JobScan provides a minimalistic yet powerful dashboard that lets employers post jobs, manage applicants, and view analytics. Recruiters define skills, compensation, job type, and role expectations — the system then parses these into embeddings that allow automated, algorithmic candidate selection. Recruiters receive AI-generated screening questions, candidate rankings, and are able to set up interview slots or offer instant micro-interviews via the Telegram bot.

The bottom layer of the stack is the Web3 integration module. This supports EVM-compatible chains such as Ethereum and Solana, enabling on-chain storage of credentials, decentralized identity (DID), and payment infrastructure. Smart contracts govern token rewards, NFT minting for achievements, staking pools for identity verification, and DAO participation mechanisms. Chainlink oracles are used for off-chain data validation, and IPFS is optionally used for decentralized resume storage. The entire backend is managed using FastAPI, PostgreSQL, Redis, and deployed in a highly available containerized infrastructure via Kubernetes and AWS/GCP.