Posted 2d ago
Computer Vision Intern (PyTorch + OpenCV)
Posted 1mo ago10+ already applied
What this role pays, in context
This listing pays 9% above the median for Data Scientist roles in India, which sits at ₹23K a month.
25th pct ₹20.3K75th pct ₹27.3K
Median of 11 other live Data Scientist listings in India on this board that state pay. Roles that do not disclose a salary are excluded rather than counted as zero.
Why this role exists
Every company that posts on MailerMen uploads a logo, a team photo or a product screenshot, and nobody checks any of it until a user complains. You will spend six months building the vision service that does the checking automatically.
The work
The first milestone is a classifier that separates real logos from placeholders, screenshots and the occasional accidental selfie. You assemble the dataset yourself from our media bucket, label it with the small annotation tool we already run, and train in PyTorch from a pretrained backbone. Then comes the harder half: detecting text inside uploaded banners so we can flag employers who hide contact details in an image to slip past our filters.
What you take away
A shipped inference endpoint, a confusion matrix you can defend in a review, and a realistic sense of how much of computer vision is data cleaning. You will be in our Baner office three days a week beside the two engineers who run the media pipeline. This suits a final-year student or recent graduate who has trained a convolutional network before and wants to try it against messy production images instead of a tidy benchmark set.
Responsibilities
- Build a labelled image dataset from the media bucket covering logos, screenshots, banners and rejected uploads
- Train an image classifier in PyTorch from a pretrained backbone to separate valid company logos from everything else
- Add a text detection step that flags contact details hidden inside uploaded banner images
- Measure per-class precision and recall, then set the confidence threshold used for auto-rejection
- Package the trained model behind an inference endpoint the upload service calls on every submission
- Track false positives reported by the moderation team and retrain once a pattern is clear
Requirements
- Final-year student or recent graduate who has trained at least one convolutional network end to end
- Python plus practical familiarity with PyTorch or an equivalent deep learning framework
- Comfortable with resizing, augmentation and colour spaces, and able to say why each matters
- Understands the trade-off between precision and recall in a moderation setting
- Available for three days a week in the Baner office
Skills
Benefits
- Monthly stipend across the full six months
- Direct mentorship from the two engineers running the media pipeline
- Ownership of a service that runs on every single upload to the platform
- GPU access for training so nothing stalls on your own laptop
- Strong interns are considered for a full-time role at the end