AI / Computer Vision Intern · Pune or remote

Who we are

We're building S18LabX, an early-stage sports-tech startup in Pune focused on sports video intelligence and AI-assisted coaching. Founded by Arjun Shah (Columbia MBA; previously at Nanonets, Credit Suisse and Barclays) and Chirag Lulla (entrepreneur and sports team owner), we aim to turn training footage into information players, coaches, and parents can use.

A coach can watch a training session and understand each action: the setup, movement, contact, outcomes, and patterns. We want AI to do the same from ordinary phone footage, despite real-world variability.

We're looking for an AI / Computer Vision Intern with strong fundamentals to work alongside our engineering team on the models and data behind the product.

This is not an internship where you'll be handed a tidy dataset and a tutorial. You'll work on the real problem with real footage, run experiments that help decide which approaches actually work, learn how we evaluate success, and see your work go into a product coaches use every day.

Responsibilities

• Run experiments on real training video with the engineering team: fine-tune vision and video models, build evaluation sets, measure results, and analyse failures
• Help develop models for tracking, action segmentation, movement, contact, and other key events in training video
• Help build the labelling tools and dataset workflows that turn coach corrections into useful training data
• Test model performance on unseen training sessions and report what works, what doesn't, and why
• Write clean, reproducible code and document each experiment so the team can build on it
• Work with the founders and engineers on experiments that shape the product
• Own a scoped problem end to end and move quickly from idea → experiment → result → next experiment

The loop you'll help build and improve

Video → AI analysis → Coach correction → Proprietary labelled data → Better models → Better product

What we're looking for

Must have

• Final-year student or recent graduate in computer science, engineering, data science, or equivalent, with strong ML fundamentals
• Hands-on experience building and training models in Python and PyTorch, from coursework or your own projects
• Working understanding of at least one area such as object detection, tracking, pose estimation, action recognition, or video understanding
• One project you built yourself end to end and can explain: what it did, how you evaluated it, and what broke
• Comfortable working with messy, noisy, real-world data and limited labelled datasets

Nice to have

• Experience with video processing tools such as OpenCV or FFmpeg
• A public GitHub, Kaggle profile, or paper that shows your work
• Broader software skills: backend, data pipelines, or application development
• Interest or experience in sports, biomechanics, or sports analytics

You do not need prior work experience or a sports background. We care more about what you've built and how you reason about problems you haven't seen before.

The kind of person who will thrive here

This is a hands-on internship on an early-stage team, not an observer's seat.

You should be comfortable taking ownership of a problem with incomplete information and no established playbook. Bring strong experimental instincts: form a hypothesis, test it quickly, measure the results, and decide what to pursue or abandon.

One week you might be fine-tuning a video model. The next, building the tool that labels a thousand training clips, or finding out why a model that worked in the notebook fails on real footage.

You'll work directly with the founders and the engineering team, and your work will reach coaches who use it every week.

This is not a research placement. We care about rigour and experimentation, alongside a strong bias toward shipping something coaches use, learning from real users, and continuously making it better.

Role

• Paid internship with a monthly stipend
• 3 to 6 months, length agreed per candidate; start date flexible
• Pune preferred; remote or hybrid considered for the right person
• Possible conversion to a full-time role