Data Scientist - Computer Vision

Overview

NowPurchase is transforming the $140B Metal Manufacturing industry. The metal industry forms the backbone of the economy and the fundamental block of the physical world - be it transportation, construction, and every machinery. NowPurchase is a rich, digital marketplace where metal manufacturers (foundries + steel plants) can procure high-quality raw materials (scrap, pig iron, ferroalloys, additives, nodularisers) in a trusted manner. Our technology allows them to optimize their manufacturing process to ensure high productivity and resilience to failure. We currently serve over 250 factories nationwide and are looking to aggressively expand our footprint across India. You can learn more on www.nowpurchase.com.

Job Description

Position – Data Scientist

Function – AI

Experience – 1 - 2 Yrs

Location – Kolkata

No. of Positions – 1no

Qualification – B.Tech in Computer Science

NowPurchase is seeking a skilled and driven Data Scientist (Computer Vision) to design, train, and deploy AI-first vision models for real-world manufacturing environments. The ideal candidate will have strong hands-on experience/knowledge with computer vision, a passion for working with production-grade data, and the ability to see their models perform on live shopfloors. This role requires a practical mindset, close collaboration with engineering and operations teams, and a strong focus on improving process efficiency, quality, and safety in foundries and metal manufacturing units.

• Translate shopfloor and business problems into clear computer vision tasks (detection, segmentation, pose, tracking, vision–language, etc.).
• Own the data lifecycle: define classes, design and perform annotations, and maintain clean, versioned datasets.
• Design and run experiments, compare models, and improve performance based on structured evaluation and error analysis.
• Prepare and optimize models for deployment on GPUs and edge devices with a focus on reliability, latency, and robustness.
• Work with developers to integrate models into production pipelines, dashboards, and alerts.
• Collaborate with AI / Product / Engineering / Design to ship solutions that are usable and valuable on the shopfloor, and document your work clearly.
  
  

• Strong programming skills in Python.
• Solid understanding of PyTorch and custom model training (beyond just running off-the-shelf notebooks).
• Practical experience with:
  
  

for object detection (training, validation, fine-tuning, augmentations, hyperparameter tuning).

RF-DETR or similar transformer-based detectors.

• Experience with segmentation using:
  
  

, YOLO-Seg

• Familiarity with pose estimation using YOLO-pose and/or other pose models (e.g., OpenPose / MMPose / RTMPose or similar).
• Basic working knowledge of VLMs like CLIP, Florence-2, and at least one other modern VLM (e.g., BLIP, LLaVA, Qwen-VL or similar).
• Exposure to edge devices like Jetson Nano/Orin or similar low-latency setups.
• Experience exporting models to ONNX, familiarity with TensorRT for optimizing and deploying models on GPUs / edge devices.
• AWS familiarity:
  
  

for datasets and model artifacts.

SageMaker (using the SageMaker SDK to launch/train/evaluate experiments).

CloudWatch for viewing logs and basic monitoring of jobs/services.

• Understanding of latency–accuracy trade-offs and how to debug performance bottlenecks.
  
  

• Compensation: As per industry standards & pedigree of the candidate
• Group Medical Insurance: This is over and above compensation. 3 lakhs floater for the family including parents, spouse, children. Top Up option is also available upon personal request.
• Generous leave structure including maternity & paternity leaves
• Snacks on the house
  
  

• Screening of applicants & telephonic discussion with HR.
• F2F/Video discussion with Hiring Managers
• Final round interview with CTO.
• Email communication on final feedback.