EducaNation.

Context

EducaNation is a full-stack web application built by a five-person team: multiple related entities, 50k+ relational records, and an interface for searching, filtering, and browsing across them. My work centered on the backend: the REST API, the data layer, and the performance of both.

Approach

RESTful API with server-side pagination and filtering. The first working version did what first versions do: shipped whole result sets to the client and let React sort it out. That collapses at 50k records. Moving pagination and query filtering into the Flask API, with SQLAlchemy composing the filters into SQL rather than Python post-processing, cut response payloads by roughly 70% and brought average API response times under 300 ms, joins and filtered search included.

Relational modeling. The entities are genuinely relational: the interesting queries are joins. Modeling them properly in PostgreSQL meant the database does what databases are good at, instead of the API reimplementing joins badly in application code.

Team workflow. Five people on one codebase works or doesn't based on process: Git-based workflows on GitLab, peer code review on every merge, iterative releases, and CI-style testing with PyTest holding backend coverage above 90%. Postman collections doubled as living documentation of the API contract.

The hard part: making the API honest under load

The tricky part of server-side filtering is that every combination of filters, sort order, and page must compose into one efficient SQL query; the naive route of fetching broadly and filtering in Python quietly reintroduces the payload problem one layer down. Getting SQLAlchemy to build the right query for every combination, and proving it with tests rather than assuming it, is where the response-time budget was actually won.

Results

• 50k+ records served across multiple entities with filtered search, joins, and pagination
• Under 300 ms average API response time
• ~70% smaller payloads after moving pagination and filtering server-side
• 90%+ backend test coverage, maintained across releases by a five-person team

Takeaways

Performance work on an API is mostly about respecting the database. And team velocity came from the boring things (review discipline, tests that fail loudly, a shared API contract) more than from any individual's code.