PromptID: AI-Native Assessment Platform Evaluates Learners' Train of Thought for Employers and Universities

PromptID is an AI-native EdTech assessment platform for employers and universities that evaluates learners by analysing their train of thought rather than rewarding memorisation, driven by a proprietary assessment algorithm and a reasoning-model evaluation path. The platform was taken from zero to production in six weeks in 2026, delivered pilot-ready and investor-demo-ready under compressed timelines. April Elias of PromptID, in a verified client review, said, "Shipped what would normally take a bigger team a lot longer. Report overhaul, security hardening, infrastructure tightening, they owned it end to end. They kept production stable through pilot prep and investor demos." The build was compressed mid-project when the market shifted and the timeline shrank by a month, with pilot conversations and investor demos already on the calendar. The team held the production-quality bar by cutting feature scope rather than testing rigor, squeezing the application surface area while keeping QA sharp. The production architecture runs a NestJS API with a BullMQ evaluation queue, a NextJS frontend, and LangChain orchestrating model-agnostic large language model calls so model swaps become a configuration change. The system deploys on autoscaling Kubernetes, with Postgres and Redis as managed services and S3 for uploads, provisioned end-to-end with Terraform. A proprietary assessment engine and reasoning-model evaluation path sit at the core. Assessments are evaluated asynchronously: a learner submits via the API, the submission is persisted and an evaluation job is enqueued, an evaluation worker invokes the reasoning model through LangChain to score the train of thought, and the learner polls to retrieve the feedback payload. This keeps the reasoning-model scoring decoupled from the submission flow. The platform's underlying mission is verified human competency: PromptID builds a tamper-proof Verified Competency Index (VCI), a portable record backed by audit-ready evidence and human oversight, positioning assessment around measured competency rather than attendance or completion. Model agnosticism and Kubernetes autoscaling proved their value, absorbing load spikes in seconds.