Checkmarked

Checkmarked creates developer portfolios where every claim is tied back to real GitHub activity. Developers connect their repos, and Checkmarked runs a two-stage analysis pipeline: a fast deterministic scan for instant feedback, followed by deeper AI passes that generate project narratives, highlights, and visuals. The platform then uses those insights to verify roles, tech stacks, and CV skills, giving recruiters and hiring teams a portfolio they can trust—complete with clear traffic-light verdicts on what the candidate actually did and knows. Under the hood, Checkmarked combines Next.js, React, and SSE-based streaming endpoints to handle long-running AI analyses without timeouts, exposing granular progress events to the UI. A unified analysis pipeline treats single- and multi-repo projects consistently while merging code context with conversation history for claim verification. AI output is constrained through Zod-validated schemas and a centralized prompting layer, which feeds into a content-block model that keeps AI-generated descriptions, roles, and impact statements synchronized with editable portfolio sections. A contribution-based verification engine inspects GitHub commits to cross-check user claims, and an Azure OpenAI integration adds error classification, exponential backoff, and schema validation to stabilize all AI-dependent workflows. For CVs, Checkmarked runs a multi-stage skill validation flow that combines deterministic GitHub heuristics, targeted code scans, and AI reasoning to classify skills as sufficient, borderline, or unsupported—with structured explanations. Client-assisted code ingestion allows candidates to submit focused code snippets within strict token budgets, turning raw code into concrete, reviewable evidence that hiring teams can trust at a glance.