- TypeScript
- MongoDB
- Express JS
- React JS
- Tailwind CSS
- Tanstack Query
Northpeak — Deadline-Aware Project Management SaaS
Northpeak is a modern, deadline-aware project management platform designed specifically for engineering teams. It keeps managers informed of delivery timelines and enables developers to organize, track, and complete granular tasks on a live Kanban board.
- ROLE
- Fullstack Developer
- DURATION
- 2 Months
- PLATFORM
- Web app

01 — Context & Problem
Context & Problem
Introduce Northpeak as a full-stack project and task management platform designed around two distinct user roles: Managers and Developers.
Explain how Managers organize projects, manage team membership, assign work, and monitor progress, while Developers focus on their assigned tasks, subtasks, and discussions
The central problem is not simply managing tasks. It is ensuring that changes to project membership and task ownership do not leave the application in an inconsistent state.
02 — Architecture
Architecture
Explain how Northpeak separates its frontend, backend, and database responsibilities. The frontend uses React 19, TypeScript, and TanStack Start, with public pages rendered through SSR and authenticated workspace routes handled separately. TanStack Query manages server state, caching, and mutations.
The backend uses Express and TypeScript, organized into routes, services, serializers, and database models. MongoDB with Mongoose provides the persistence layer for users, projects, tasks, subtasks, comments, activity, and notifications.
Data Relationships
The diagram shows logical relationships represented by ObjectId fields. These are not relational-database foreign keys. MongoDB does not automatically prevent a referenced document from being deleted.
03 — Engineering Decisions
Engineering Decisions
Explain the important technical decisions as part of the engineering narrative, rather than presenting them as a list of technologies.
DECISION
— Assignment Consistency
Northpeak needed to handle situations where project membership changes could conflict with task assignments. Instead of relying on a single update operation, the implementation uses optimistic concurrency, database transactions, post-write verification, compensating actions, and reconciliation to protect assignment consistency.
DECISION
Database Architecture
MongoDB with Mongoose was selected because the application contains several evolving entities, including projects, tasks, subtasks, comments, and activity records. The flexible document model suited the evolving application structure, while relationships and consistency across documents required deliberate engineering safeguards.
DECISION
Authentication
JWT authentication was implemented for communication between the separate frontend and backend. localStorage was initially selected as the token-persistence mechanism in the traditional client-server architecture. The authentication lifecycle subsequently incorporated OTP verification, Google OAuth, database-backed role verification, and session-aware cache handling.
04 — Challenges & Solutions
Challenges & Solutions
Focus this section on the engineering problems that required more than ordinary feature implementation.
The most significant challenge involved concurrent project membership changes and task assignments. The implementation coordinates these operations through optimistic concurrency and transactions, then verifies the resulting state and uses compensating actions when necessary. A reconciliation mechanism provides an additional recovery path.
Another challenge involved testing concurrency without relying on unpredictable delays. The project introduces deterministic concurrency barriers that allow tests to control the order of operations. These barriers are restricted to the test environment, helping the concurrency tests verify specific race conditions without introducing production-only testing hooks.
Authentication also required careful handling. OTP verification uses atomic attempt limits and timing-safe comparison, while account creation is protected against duplicate registration. Google OAuth account linking includes safeguards against an unverified account taking over an existing account.
On the frontend, excessive refetching, loader flicker, and potential cross-user cache exposure were addressed through tiered query freshness, controlled loading behavior, and authentication-aware cache invalidation.
Outcomes
05 — Trade-offs & Reflection
Trade-offs & Reflection
Conclude by discussing what the implementation does not yet solve and how the architecture could evolve.
JWT persistence through localStorage was a practical initial choice, but it introduces token exposure risks if an XSS vulnerability occurs. HTTP-only cookies would be worth evaluating as an alternative. Similarly, the current in-memory rate-limiting approach would need a shared store such as Redis if the application required consistent rate limiting across multiple instances.
