Computer Vision Architect

Infosys; Bengaluru; On-site; INR 4,000,000 - 6,000,000; Full time

Job summary

Infosys is seeking a Computer Vision Architect to design, train, and optimize deep learning models for real-time video analytics, including object detection, recognition, and segmentation.

You will lead large-scale deployments across 100+ streams, fine-tune CNNs/ViTs with TensorFlow, PyTorch, and ONNX, and build inference pipelines using NVIDIA tools.

Qualifications

• Experience in building CV models for real-time video analytics.
• Proficiency in CNNs, Vision Transformers, and related DL frameworks.
• Experience with NVIDIA tools such as DeepStream, TAO Toolkit, and Triton.

Responsibilities

• Lead real-time multi-stream video processing at scale.
• Fine-tune CNNs and Vision Transformers with TensorFlow, PyTorch, and ONNX.
• Develop inference pipelines using NVIDIA tools such as NGC models and Triton Server.
• Apply pre- and post-processing to optimize model performance.
• Implement multi-object tracking across camera feeds and perform ROI extraction.

Skills

Feature extraction

Latent space analysis

Data manipulation

CV/ML model optimization

Tools

Terraform

Ansible

Docker

Kubernetes

Neo4j

SQL

NoSQL

ONNX

TAO Toolkit

DeepStream

Triton Server

Job description

Computer Vision Architect Design, train, and optimize deep learning models for computer vision tasks such as object detection, image recognition, segmentation, and video analytics.

Key Responsibilities

• Lead large-scale real-time implementations involving 100+ video streams.
• Fine-tune models like CNNs and Vision Transformers using frameworks such as TensorFlow, PyTorch, and ONNX.
• Build and deploy inference pipelines using NVIDIA technologies including NGC models, TAO Toolkit, DeepStream Metropolis, and Triton Server.
• Apply pre- and post-processing techniques to enhance model performance and evaluation.
• Develop and implement multi-object tracking algorithms across multiple camera feeds., Perform camera calibration and extract regions of interest (ROI), action recognition using temporal analysis techniques.

Skills and Qualifications

• Develop models for feature extraction and come up with strategies for latent space analysis and manipulation of the data.
• Proficiency in writing complex SQL queries, familiarity with NoSQL or Graph Databases (Neo4j) is an advantage.
• Knowledge of infrastructure as code tools (e.g., Terraform, Ansible) is an advantage.
• Proficiency with containerization technologies (Docker) and orchestration tools (Kubernetes) is an advantage.