neeJou

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for neeJou

Inspiration

Hiring software engineers from a resume is noisy. A resume can say "PHP", "AI", or "e-commerce", but it cannot prove whether someone has actually shipped a similar system. neeJou started from a simple belief: we do not read resumes, we read Git.

The goal is to help project owners find engineers through real implementation evidence. If an engineer has built a repository whose structure, stack, commits, README, and system shape overlap with a client's project, that is stronger evidence than a polished profile. In some cases, that overlap can even reduce delivery cost because existing patterns can be migrated instead of rebuilt from zero.

What it does

neeJou is an AI-powered project-to-engineer matching platform.

On the client side, a project owner describes what they want to build. The assistant turns that loose request into a structured project profile: project type, language, database, budget, cross-border preference, and requirement notes. When the client explicitly asks to start matching, neeJou searches for candidate engineers.

On the engineer side, users sign in with GitHub and add repositories. neeJou verifies that the repository belongs to the signed-in GitHub user, checks detected languages, reads repository metadata, samples recent commit activity, inspects the file tree, extracts README context, and stores a repo profile snapshot.

The matching step first filters by hard constraints such as category, language, database, budget, country preference, and repository quality gate. Then AI evaluates candidate repositories semantically and recommends the best fit with a score, reason, GitHub project, and contact card.

How we built it

The prototype is a PHP/MySQL web app with Google login for clients and GitHub OAuth for engineers. GitHub API calls collect repository metadata, languages, README content, file paths, dependencies, and recent commit details. A scoring layer evaluates stack match, activity, implementation substance, project origin, and repository health.

OpenAI Responses API powers the project setup assistant and the repository evaluation phase. The assistant uses tool calls to update structured project details and only runs matching after explicit user intent such as "start matching". The repository evaluator compares candidate repo snapshots against the client project and returns one practical recommendation.

Codex was used heavily to inspect the legacy PHP codebase, trace the matching flow, identify where the product story lived, implement and validate the repo snapshot pipeline, and turn a rough prototype into a coherent hackathon submission. Codex helped keep the focus on a working evidence chain instead of adding unrelated marketplace features.

Challenges we ran into

The hardest part was avoiding a shallow keyword matcher. Matching "PHP + MySQL" is easy, but it is not enough. We needed the system to look at repository shape, commit behavior, README signals, and whether the project appears to contain real implementation work. Another challenge was keeping the AI bounded: the assistant should stabilize project requirements first, and it should not claim a match unless the actual matching tool ran.

Accomplishments that we're proud of

We built an end-to-end evidence chain: client requirements become structured data, engineer repositories become scored snapshots, and matching is based on repository evidence rather than self-promotion. The demo can show both sides of the marketplace and explain why a specific engineer is recommended.

What we learned

Repository evidence is much richer than a resume, but it has to be structured before AI can reason over it reliably. The best pattern was a two-stage flow: deterministic filters and quality gates first, AI semantic evaluation second. That keeps the result explainable while still allowing nuanced judgment.

What's next for neeJou

Next steps include deeper code intelligence, private repository support with explicit consent, migration-cost estimates, better anti-fraud checks, and a cleaner hosted demo environment for project owners and engineers.

Built With

• ai-agents
• codex
• developer-tools
• github-api
• javascript
• mysql
• openai
• php
• project-matching
• responses-api
• tailwind-css