Deploy IndicTrans2 Translation API

@aurvixlabs

This is exactly the kind of focused deployment work my team and I enjoy: you already have the Linux box, you know the repos you want (IndicTrans2 and dl-translate), and you just need a clean, repeatable setup that “just works” behind a simple HTTP endpoint. The success criteria you’ve outlined are clear—one command to start the service, a curl call that returns a valid Hindi translation, and the ability to switch between both model backends via the same endpoint. I’m an AWS Machine Learning Specialty certified engineer leading a small, experienced team, and we regularly turn research-grade repos into stable services with straightforward start/stop flows and concise READMEs. We’ll set up the environment, preload a couple of useful language pairs, and give you either systemd or Docker/compose scripts so you can bring the service up the same way every time. Looking forward to message in your chat Tanay Kapoor Aurvix Labs [login to view URL]

@ravikumarsingh1

I can set up a clean, reproducible Linux deployment of both IndicTrans2 and dl-translate, exposed through a single lightweight REST API that you can call directly with curl. What I’ll do • Recommend and configure the best-fit Python 3.x environment • Install all dependencies and download required models for IndicTrans2 and dl-translate • Build a minimal REST service (FastAPI or Flask) with one endpoint: POST /translate → accepts source text, source language, target language, and model selector • Ensure both translation engines load correctly and can be swapped via the same endpoint • Preload a small set of demo language pairs while keeping the setup extensible Operations & reproducibility • Provide Dockerfile + docker-compose or systemd start/stop scripts (your choice) • Service starts with a single command • Clean logging and predictable startup behavior Deliverables • Working translation service callable via curl • Start/stop mechanism (Docker or systemd) • Short README with install steps and example curl commands Acceptance • One-command startup • curl -X POST http://localhost:port/translate returns valid translations • Both models load without errors and function through the same API I’m comfortable with ML model deployments on Linux and will keep the setup simple, repeatable, and easy to extend later.