Vision AI · Nashik (HQ, Maharashtra) · Full-time
Senior Computer Vision Engineer
Build and ship the models that score store-walk photos against each franchisor's brand manual: signage, menu boards, planograms, cleanliness and equipment condition.
Gazi Wellness's Vision AI looks at photos taken by Franchise Business Consultants and store managers and checks them against brand standards. A typical photo is taken on a mid-range phone, under fluorescent light, at an angle, in a kitchen or a parking lot. The model has to say whether the pylon sign has a dead LED module, whether the menu board shows the current promotion, or whether the prep line is clear, and it has to explain itself well enough that a franchisee accepts the finding.
You will join a team of five led by our CTO, Priyanka Menon. Every new customer brings a new brand manual, so most of the work is making the models adapt to a franchisor's reference photos quickly and with few examples, while keeping false positives low enough that FBCs trust the output.
What you will do
- Own detection and classification models for exterior signage, POP and campaign materials, menu boards and back-of-house cleanliness.
- Design few-shot and fine-tuning workflows that adapt models to a new franchisor's approved reference photos in days, not weeks.
- Build evaluation sets with our field operations team and report precision and recall per rubric item, per customer.
- Ship models to production on AWS, including batch scoring after offline sync and low-latency scoring for on-site reports.
- Work with the mobile team on on-device image quality checks (blur, glare, framing) before upload.
- Review annotation guidelines and spot-check labeling quality.
What you bring
- 5+ years building computer vision systems, with at least 2 years shipping models to production.
- Strong Python and PyTorch; experience with object detection and image classification architectures.
- Experience with few-shot learning, transfer learning or vision-language models in a product setting.
- Comfort designing evaluation sets and explaining model errors to non-ML colleagues.
- Experience with model serving and monitoring (for example, containerized inference on AWS).
Nice to have
- Retail shelf recognition, planogram compliance or OCR on signage.
- Experience with on-device models for iOS or Android.
- Familiarity with restaurant or retail operations.