Diploma in Model Compression (Quant/Prune/Distill)
Online postgraduate diploma of 800 hours in Model Compression (Quant/Prune/Distill): 8 modules covering the computational cost of neural networks, quantisation, pruning, knowledge distillation, factorisation and efficient design, compression tools and workflows, and evaluation of compressed models, with an assigned technical tutor.
Diploma in Model Compression (Quant/Prune/Distill) — studying from Australia
Electronics and embedded systems underpin a wide range of Australian industries, including defence, mining automation, medical devices and consumer technology. The telecommunications sector is significant, with ongoing investment in fibre, mobile networks and satellite communications for remote regions. Research organisations such as CSIRO and university groups contribute to advances in sensing, communications and signal processing. Engineers typically work on hardware design, firmware development, testing and certification. Employers include telecommunications carriers, defence contractors, instrumentation companies and product manufacturers.
Fees for Australia: A$2,380 (AUD), with instalments and employer sponsorship available.
Relevant bodies in Australia
Typical roles
Questions from Australia
Is this qualification recognised for electronics engineering in Australia?
SEIUM's qualifications are private and not official state degrees, and they are not accredited by Engineers Australia or the Australian Communications and Media Authority. Recognition for electronics and telecom roles generally depends on accredited degrees, licences and experience. Employers or the relevant body make the final decision.
Which Australian employers hire embedded and telecom engineers?
Australia has telecommunications carriers, defence electronics firms, medical device makers, mining technology companies and IoT developers. Employers mostly assess your hands-on skills, project record and certifications. Ask prospective employers whether they would consider a private online qualification.
Can I study this part-time while working, and how is it supported?
The courses are 100% online and taught in English, so you can study from anywhere in Australia without needing a student visa. Specialist courses take 300 hours, diplomas 800 and master's 1,900. Each student has an assigned technical tutor, fees are payable in instalments, and employer sponsorship may be possible.
Take the controls.
Run a real engineering trade-off from this field. Move the sliders, watch the model react in real time and beat the challenge.
What you will be able to do
Analyse the computational and memory cost of neural networks.
Apply post-training and quantisation-aware training techniques.
Reduce models through structured and unstructured pruning.
Transfer knowledge from large models through distillation.
Use factorisation and efficient architectures.
Automate compression workflows with industry tools.
Evaluate compressed models for accuracy, latency and robustness.
Who it's for
- Machine learning engineers who need to deploy models under tight latency or memory budgets.
- Model optimisation engineers working with quantisation, pruning and distillation.
- Embedded AI engineers bringing neural networks to NPUs and microcontrollers.
- Researchers in neural network efficiency seeking hands-on industrial workflows.
- MLOps engineers automating compression in the deployment pipeline.
- Edge AI solution architects choosing models and hardware for on-device inference.
- Engineering or science graduates looking to enter the consumer electronics sector.
What you'll work with
- ISO/IEC 15938-17:2022 (compression of neural networks for multimedia content description and analysis)
- ISO/IEC 23053:2022 (framework for AI systems using machine learning)
- IEEE 754-2019 (floating-point arithmetic)
- Regulation (EU) 2024/1689 – AI Act (harmonised rules on artificial intelligence)
- ISO/IEC 22989:2022 (AI concepts and terminology)
- ISO/IEC 42001:2023 (AI management systems)
- PyTorch (torch.ao)
- TensorFlow Model Optimization Toolkit
- NVIDIA TensorRT Model Optimizer
- Intel Neural Compressor
- Hugging Face Optimum
8 modules
1.1Parameters, FLOPs and MAC operations1.2Weight and activation memory1.3Compute versus bandwidth: the roofline model1.4Efficient architectures: MobileNet and EfficientNet1.5Transformers and the cost of attention1.6Latency, throughput and energy metrics1.7Target hardware: CPU, GPU, NPU and MCU1.8Trade-off between accuracy and size1.9Reference benchmarks such as MLPerf1.10Lab: profiling with PyTorch Profiler
2.1INT8, INT4, FP16 and FP8 representations2.2Symmetric and asymmetric quantisation2.3Per-tensor and per-channel granularity2.4Calibration: min-max, entropy and percentiles2.5Post-training quantisation (PTQ)2.6Quantisation-aware training (QAT)2.7Fake quantisation and the straight-through estimator2.8LLM quantisation: GPTQ and AWQ2.9Typical errors and sensitive layers2.10Practical: PTQ and QAT of a ResNet in INT8
3.1Unstructured magnitude pruning3.2Structured pruning of channels and filters3.3The lottery ticket hypothesis3.4Iterative pruning and retraining3.52:4 sparsity on NVIDIA GPUs3.6Gradient-based importance criteria3.7Pruning attention heads3.8Real versus theoretical speed-up3.9Combining pruning and quantisation3.10Lab: structured pruning of a CNN
4.1Teacher model and student model4.2Soft-target loss and temperature4.3Distillation of intermediate features4.4Attention and relational distillation4.5Self-distillation and online distillation4.6Distillation of language models4.7Choosing the student architecture4.8Synthetic data for distillation4.9Evaluating knowledge transfer4.10Workshop: distilling BERT into a smaller model
5.1Low-rank decomposition using SVD5.2Tensor factorisation: CP and Tucker5.3LoRA adapters and their merging5.4Weight sharing and hashing5.5Neural architecture search (NAS)5.6Depthwise separable convolutions5.7Early exit and dynamic computation5.8Mixture of experts and sparsity5.9Compression of embedding tables5.10Practical: low rank in linear layers
6.1PyTorch quantisation API6.2TensorFlow Model Optimization Toolkit6.3Quantisation with ONNX Runtime6.4NVIDIA TensorRT Model Optimizer6.5Intel Neural Compressor and OpenVINO NNCF6.6Hugging Face Optimum6.7Export and operator compatibility6.8Automating the compression pipeline6.9Model version control6.10Lab: compression pipeline with NNCF
7.1Accuracy metrics after compression7.2Latency on the target device7.3Robustness and bias after compression7.4Layer-wise sensitivity analysis7.5Confidence calibration7.6A/B testing in production7.7Model documentation and model cards7.8Requirements of the EU AI Act7.9Selecting the optimal strategy7.10Case study: compressing a YOLO detector for the edge
8.1Selection of the model and hardware8.2Accuracy and latency baseline8.3Applying quantisation8.4Combined pruning and distillation8.5Export to the target runtime8.6Comparative evaluation of variants8.7Trade-off analysis and recommendations8.8Technical report on the model8.9Report and presentation8.10Final project: final review with your tutor
How you'll learn
The programme is delivered 100% online on the SEIUM campus, with downloadable material, recorded classes and an assigned technical tutor. There are no fixed timetables: you progress at your own pace within the intake, so you can combine it with your job.
Each module combines theory, guided reading of the original standard and applied exercises. The working software is PyTorch (torch.ao), TensorFlow Model Optimization Toolkit, NVIDIA TensorRT Model Optimizer, Intel Neural Compressor and Hugging Face Optimum, in free, academic or evaluation versions. Testing is studied through real procedures: quantisation and pruning of vision models, latency measurement on edge devices and accuracy comparison after compression. Assessment is continuous, module by module, plus the final project.
Standards covered: ISO/IEC 15938-17:2022 (neural network compression), ISO/IEC 23053:2022, IEEE 754-2019, Regulation (EU) 2024/1689 (AI Act), ISO/IEC 22989:2022 and ISO/IEC 42001:2023.
Build it end to end
Your final project follows the real engineering workflow, step by step, reviewed by your tutor.
Where this takes you
Requirements
- Prior technical training or professional experience in the sector. No university degree is required.
- An internet connection and a computer capable of running the programme software.
- English sufficient to read technical standards.
Ready to start?
Send your application today. Our admissions team replies within 72 hours with your study plan and funding options.
Frequently asked questions
The programme totals 800 hours over 6–8 months. You set your own pace within each intake, and your tutor helps you plan the workload around your job.
You work with PyTorch (torch.ao), TensorFlow Model Optimization Toolkit, NVIDIA TensorRT Model Optimizer, Intel Neural Compressor and Hugging Face Optimum, using free, academic or evaluation versions. Your tutor indicates the free alternative wherever one exists.
A SEIUM Institute diploma, awarded on passing the continuous assessment and the final project. It is SEIUM's own qualification, not an official state degree; its recognition depends on the programme completed and the regulations of each country.
Yes. Modules passed are recognised in a higher programme in the same area (postgraduate diploma or master's). Ask about the recognition policy when you enrol.
Prior technical training or professional experience in the sector (no university degree required), a computer able to run the programme software and English sufficient to read technical standards.
Start dates are set by SEIUM's faculty for each intake. Our admissions team will confirm the next available date when you apply.
Related programmes
Diploma in Applied Cryptography and Anti-rollback
Online postgraduate diploma of 800 hours in applied cryptography and anti-rollback. 8 modules covering cryptographic primitives, embedded cryptography, key management and PKI, firmware signing, anti-rollback protection,…
Diploma in DFM/DFT and Process Control at EMS
Online postgraduate diploma of 800 hours in DFM/DFT and process control in EMS. 8 modules covering EMS manufacturing, design for manufacturing, design for test, printing and…
Postgraduate Diploma in IEC/UL/EN and CE Marking
Online 800-hour Postgraduate Diploma in IEC/UL/EN and CE Marking. 8 modules covering Directive 2014/35/UE, Directive 2014/30/UE and Rohde & Schwarz ELEKTRA. Assigned tutor and open enrolment.
More in Electronics, Embedded Systems & Telecom
All Electronics, Embedded Systems & Telecom programmes → · How to apply · Fees & financing · Scholarships