Prognostix

Predicts how long a jet engine has left

Demo — deploying· Render (Docker)Read the sourceMostly Jupyter Notebook

Results — hover any figure for the baseline and the caveat

Lifecycle stages, one repo

Ingest, train, evaluate, serve, monitor — with a drift check and alerting. Model performance numbers aren't published yet; that's the next thing this repo needs.

Overview

A predictive-maintenance platform that estimates the remaining useful life of industrial equipment from sensor telemetry, built on NASA's CMAPSS turbofan degradation data. The reason it's here isn't the model — it's everything around it: a feature pipeline, an inference API, drift monitoring, and a Docker Compose stack that brings the whole thing up in one command.

Run it yourself

source

5 commands

git clone https://github.com/HasnatKhan010/prognostix-ml.gitcd prognostix-ml# models are bind-mounted, not baked into the image:python scripts/prepare_data.pypython scripts/train.py --model grucd docker && docker compose up --build

Dashboard localhost:8080 · API docs localhost:8000/docs

4 steps

01Clone the repo.
02Build the dataset, then train one model.
03Bring the stack up with Compose.
04Open the dashboard and run an inference.

Needs Docker with Compose v2 · Python 3.10+ for the training step · ~6 GB disk

No published image and no release — you build it. Skip the training step and the stack still comes up, but /health reports “degraded” and inference 503s, because the weights live in a mounted volume rather than in the image.

Built into it

LSTM, GRU and attention architectures against a Random Forest baseline, so the deep models have something to beat
Rolling and lag window feature generation as a scripted pipeline, not notebook cells
FastAPI inference service with automatic Swagger docs
Drift and performance-degradation monitoring with alert generation
docker-compose brings up API and frontend together; pytest covers preprocessing, features and the API

Built with

PyTorchFastAPIDockerpytest

Repository

HasnatKhan010/prognostix-ml

Last push 3 Sept 2026

Next projectDataPrep Studio Back outAll 26 projects

Overview

git clone https://github.com/HasnatKhan010/prognostix-ml.gitcd prognostix-ml# models are bind-mounted, not baked into the image:python scripts/prepare_data.pypython scripts/train.py --model grucd docker && docker compose up --build

Built into it

LSTM, GRU and attention architectures against a Random Forest baseline, so the deep models have something to beat

Rolling and lag window feature generation as a scripted pipeline, not notebook cells

FastAPI inference service with automatic Swagger docs

Drift and performance-degradation monitoring with alert generation

docker-compose brings up API and frontend together; pytest covers preprocessing, features and the API