Machine Learning Intern
Atlys
Software Engineering
Delhi, India
Posted on Apr 7, 2026
Job description
🎯 Atlys' mission is to enable every person on earth to travel freely.
At Atlys, we believe that the path to creating a more open world is by making it efficient to travel. Travelers cite visas as the most frustrating pain point, and we're starting by automating that completely. We're looking for talented people who are interested in building the future of travel alongside us.
Building technology to increase global movement liquidity will be one of the most exciting developments in decades. If you are curious why the smartest people want to work at Atlys, read this post.
We're looking for an ML intern to work on image quality detection and structured data extraction models. Here, you'll be building models from scratch, running real experiments and improving them through better data and smarter training techniques. You'll ship work that directly impacts production.
Job requirements
The Job:
• Train and finetune models for image quality detection and enhancement tasks
• Run structured experiments - compare different architectures, track what works and document why
• Prepare and validate labeled datasets for structured field extraction (names, dates, numbers and more)
• Write scripts to automate repetitive data processing tasks so you're not doing the same thing twice
• Document your training procedures and results clearly
The Ideal Candidate:
• Good Python skills - you can write clean, working scripts without hand-holding
• Familiarity with PyTorch and/or TensorFlow - you've trained models before, whether for coursework, competitions, or personal projects
• Understanding of computer vision fundamentals - image processing, augmentation and common architectures (YOLO, U-Net, ResNet, etc.)
• Attention to detail - data work can be tedious and mistakes propagate fast, you need to care about getting it right
• Familiarity with Hugging Face Transformers and model hubs
• Curiosity about the LLM/VLM space and willingness to learn finetuning techniques (LoRA, QLoRA, etc.)
or