Go back to projects

Research Infrastructure & APIs

#Full Stack #FastAPI #React #Docker #Infrastructure

Overview

Built the bridge between research models and end users at CERIST: APIs to serve trained NLP models, interfaces to visualize and interact with results, and the deployment infrastructure to run all of it reliably.

Engineering Stack

  • Backend: FastAPI, Django REST Framework, exposing model inference as REST endpoints.
  • Frontend: React, TypeScript, for interactive result visualization and internal tooling.
  • Infrastructure: Docker for containerized, reproducible deployment; PostgreSQL and MongoDB for structured and semi-structured data management.
  • DevOps: Administration of Linux servers and GPU workstations, including environment setup for training and inference workloads.

Why this matters

A recurring gap between NLP research and usable systems is deployment: a model that only runs in a notebook has no value to end users. This project’s goal was to make every research model in the group usable as a real, queryable service, not just a checkpoint file.

My Role

Owned the infrastructure end to end: API design, containerization, deployment, and the Linux/GPU environment the models actually train and run on.