Deep Learning Engineering Intern — Edge Vision

Role Overview

InSiSo Technologies is looking for a Deep Learning Engineering Intern — Edge Vision to work on InSiSoNet, a proprietary, hardware-agnostic vision AI architecture. This is a 3-month paid internship with a fast-track to full-time for strong performers. Candidates must be willing to relocate to Bengaluru for full-time roles and client deployments.

Responsibilities

• Improve detection accuracy for industrial and logistics use cases including object detection, quality inspection, and anomaly detection.
• Train and evaluate models on the InSiSoNet pipeline.
• Perform INT8 quantisation and quantisation-aware training to reduce size and latency.
• Handle ONNX export and compilation for edge accelerators.
• Analyse failure modes on real customer imagery and implement fixes.
• Contribute to internal annotation and training tooling.
• Document work for reproducibility.

Requirements

• Strong grasp of object detection architectures (anchor-free heads, feature pyramids, loss design).
• Deep expertise in PyTorch (custom layers, training loops, and loss functions).
• Experience in model optimisation: quantisation, pruning, and knowledge distillation.
• Proficiency in ONNX export and debugging.
• Ability to implement research papers.
• Strong Python, Git, and Linux skills.
• Numerical intuition and ability to interpret PR curves and confusion matrices.

Nice to Have

• Experience training models from scratch.
• Knowledge of inference runtimes like TensorRT, OpenVINO, TFLite, or Hailo SDK.
• Efficient architecture design for constrained hardware.
• C++ programming skills.
• Published research, Kaggle rankings, or open-source contributions in Computer Vision.

Skills

• PyTorch
• Python
• ONNX
• Computer Vision
• Linux