Coding Assessment Platform with Custom Compiler Engine

The Problem

An EdTech company was using a third-party service for their coding assessments. It worked, but the costs were climbing and they had no control over the execution environment. They wanted to bring the entire assessment platform in-house, from the student-facing UI to the code execution engine.

The scope was bigger than a typical build: the platform needed real-time code compilation and execution across multiple programming languages, with test case validation, scoring, and a clean student interface.

What We Built

A complete coding assessment platform with three major components:

• Student portal: a clean interface for taking coding assessments, with a built-in code editor, timer, and real-time feedback
• Admin module: for creating assessments, managing test cases, and reviewing results
• Custom compiler engine: the core technical achievement. An in-house code execution service that compiles and runs student code against test cases in a sandboxed environment

The compiler engine was built under direct guidance from our CTO when the team discovered that the third-party code execution service (OneCompiler) was too expensive at scale. Instead of accepting the cost, the team built their own, a decision that saved the client significant ongoing operational costs.

Tech Stack

• React + Vite + TypeScript: fast, type-safe frontend for the student and admin portals
• Firebase Cloud Functions + Node.js: serverless backend handling assessment logic, scoring, and code execution orchestration
• Firebase RTDB + Firestore: real-time database for live assessment state, Firestore for structured data
• Custom compiler engine: sandboxed code execution replacing OneCompiler, built to handle multiple languages at scale

How We Worked

This project started as a web application build (1 engineer) and expanded into a full assessment platform (3 engineers + 1 PM), a natural evolution as the client saw the quality of the initial delivery and expanded scope.

The compiler engine was the hardest piece. Building reliable, sandboxed code execution from scratch is non-trivial, security, timeout handling, multi-language support, and output validation all needed to work flawlessly for a production assessment platform.

The Result

The admission module is live and in production. The assessment module is in active development. The client is a repeat client: they started with the web app and came back for the full assessment platform, which speaks to the quality of the initial delivery.