BAMU · Second Year B.E./B.Tech Artificial Intelligence / AI & ML / AI & DS · NEP 2020
Applied Accelerated AI
HT-2 · Honours (Semester IV) · Semester IV · 3 credits · CIA 40 + SEE 60
Only for the Honours degree, open to students with a first-year CGPA of 7.5 or more.
Module 1: Introduction to Accelerated AI
15 hoursWhy AI needs acceleration — training versus inference bottlenecks; CPUs, GPUs, TPUs and NPUs; CUDA, cuDNN, TensorRT and OpenCL; and frameworks that use acceleration: TensorFlow, PyTorch and ONNX.
Examined in the 60-mark end-semester paper
- Training versus inference bottlenecks
- CPUs, GPUs, TPUs and NPUs
- General-purpose versus task-specific compute
- CUDA, cuDNN and TensorRT
- Accelerated AI frameworks: TensorFlow, PyTorch and ONNX
Module 2: Model Optimisation and Deployment
15 hoursPruning, quantisation and distillation; converting models to ONNX and optimisation pipelines; GPU inference with TensorRT; acceleration on edge devices such as Jetson Nano, Coral and RKNN; real-time vision, speech and recommendation applications.
Examined in the 60-mark end-semester paper
- Model compression for the edge
- Quantisation: FP32 to INT8
- Knowledge distillation
- ONNX conversion and optimisation pipelines
- GPU inference with TensorRT
- Edge AI devices: Jetson, Coral and RKNN
Module 3: Case Studies and Industry Applications
15 hoursAccelerated AI in autonomous vehicles, healthcare and robotics; distributed and parallel training with Horovod or PyTorch DDP; real-time video analytics; edge–cloud collaborative AI; power, deployment and latency trade-offs.
Examined in the 60-mark end-semester paper
- Accelerated AI applications
- Distributed training with Horovod and PyTorch DDP
- Real-time video analytics
- Edge-cloud collaborative AI
- Power, latency and deployment trade-offs
Module contents are summarised from the official BAMU syllabus . AiCanCode.org is not affiliated with Dr. Babasaheb Ambedkar Marathwada University (BAMU).