Diploma in On-Device Pipeline Design with TFLite/ONNX
Online postgraduate diploma of 800 hours in on-device pipeline design with TFLite/ONNX. 8 modules covering on-device inference, model conversion and export, acceleration and delegates, pre- and post-processing, pipeline orchestration, testing and fleet deployment, with an assigned technical tutor.
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
Identify the requirements and constraints of on-device inference.
Convert and validate models in TFLite and ONNX formats.
Accelerate inference with specific delegates and runtimes.
Implement efficient and consistent pre- and post-processing.
Orchestrate multi-model pipelines in real time.
Test quality and performance on real devices.
Deploy and maintain models across device fleets.
Who it's for
- On-device machine learning engineers taking models from the notebook to the device.
- Vision application developers building inspection and detection apps.
- Edge systems engineers working with Jetson, Raspberry Pi and industrial PCs.
- AI-focused mobile developers integrating models into Android and iOS apps.
- Computer vision engineers optimising real-time pipelines.
- IoT solutions architects designing distributed inference systems.
- Engineering or science graduates looking to move into manufacturing.
What you'll work with
- ONNX 1.x (Open Neural Network Exchange)
- Khronos NNEF 1.0 (Neural Network Exchange Format)
- ISO/IEC 23053:2022 (framework for AI systems using ML)
- Regulation (EU) 2016/679 (GDPR)
- Regulation (EU) 2024/2847 (Cyber Resilience Act)
- ISO/IEC 25010:2023 (software product quality model)
- TensorFlow Lite (LiteRT)
- ONNX Runtime
- NVIDIA TensorRT and DeepStream
- Google MediaPipe
- Intel OpenVINO
8 modules
1.1Advantages: latency, privacy and cost1.2Target devices: mobiles, Jetson and Raspberry Pi1.3Stages of an on-device pipeline1.4Memory, energy and temperature constraints1.5Model formats: TFLite, ONNX and TorchScript1.6Comparison of runtimes1.7Industrial and field use cases1.8Non-functional requirements1.9Reference architecture1.10Workshop: requirements analysis for an inspection app
2.1Converting Keras to TFLite2.2Exporting PyTorch to ONNX2.3Opsets and operator compatibility2.4Dynamic and static shapes2.5Graph optimisation with onnx-simplifier2.6Conversion between ONNX and TFLite2.7Numerical validation after conversion2.8Model metadata2.9Troubleshooting common errors2.10Practical exercise: exporting a YOLO detector to ONNX
3.1TFLite delegates: GPU, NNAPI and XNNPACK3.2ONNX Runtime execution providers3.3TensorRT on NVIDIA Jetson3.4OpenVINO on Intel CPU and iGPU3.5Core ML on Apple devices3.6Quantisation for accelerators3.7Threads and CPU affinity3.8Benchmarking with official tools3.9Selecting the optimal backend3.10Lab: comparison of delegates on Android
4.1Image, audio and signal capture4.2Resizing, normalisation and colour formats4.3Pre-processing on the GPU and within the graph4.4Windowing of streaming signals4.5Non-maximum suppression4.6Decoding boxes and masks4.7Object tracking across frames4.8Temporal smoothing of predictions4.9Consistency between training and inference4.10Practical exercise: vision pipeline with NMS post-processing
5.1Pipelines with MediaPipe5.2GStreamer and NVIDIA DeepStream5.3Chaining several models5.4Queues, buffers and concurrency5.5App life-cycle management5.6Energy saving and scheduling5.7Error handling and degraded mode5.8Local logging and synchronisation5.9User interfaces in the field5.10Lab: multi-model pipeline with DeepStream
6.1Unit tests for pre- and post-processing6.2Parity tests against the reference model6.3On-device test sets6.4Performance and thermal testing6.5Testing across different device models6.6Robustness testing: lighting, noise and occlusion6.7Continuous integration with device farms6.8Quality metrics in the field6.9Pipeline validation report6.10Case study: validation of a visual inspection app
7.1Packaging the model in the application7.2Dynamic model download7.3Model versioning and compatibility7.4On-device A/B testing7.5Privacy-respecting telemetry7.6Federated learning: concepts7.7Protecting the model against extraction7.8Secure model updates7.9Total cost of operation7.10Workshop: model update strategy for a fleet
8.1Use case and device8.2Starting model8.3Conversion and optimisation8.4Pre- and post-processing8.5Pipeline orchestration8.6Testing and validation8.7Deployment strategy8.8Final on-device performance evaluation8.9Report and presentation8.10Final project: final review with the tutor
How you'll learn
Training is 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 the programme fits around your job.
Each module combines theory, guided reading of the original standard and applied exercises. The working software is TensorFlow Lite (LiteRT), ONNX Runtime, NVIDIA TensorRT and DeepStream, Google MediaPipe and Intel OpenVINO, in free, academic or evaluation versions. Exercises are based on real procedures: benchmarking runtimes and delegates on different devices, numerical parity testing after conversion and field validation of inspection applications. Assessment is continuous by module, plus the final capstone project.
Standards covered: ONNX 1.x, Khronos NNEF 1.0, ISO/IEC 23053:2022, Regulation (EU) 2016/679, Regulation (EU) 2024/2847, ISO/IEC 25010:2023.
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 TensorFlow Lite (LiteRT), ONNX Runtime, NVIDIA TensorRT and DeepStream, Google MediaPipe and Intel OpenVINO, 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 is required.
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 Mobility Policy and KPI Evaluation
Online postgraduate diploma of 800 hours in mobility policy evaluation and KPIs: 8 modules covering public mobility policy, indicator systems under ISO 37120 and ISO 37122,…
Diploma in Multimodal Model Evaluation and Robustness
Online postgraduate diploma of 800 hours in multimodal model evaluation and robustness: 8 modules covering multimodal models, evaluation design, performance according to ISO/IEC TS 4213, robustness…
Diploma in Operational Metrics and Low Latency KPIs
Online postgraduate diploma of 800 hours in operational metrics and low-latency KPIs. 8 modules covering KPI design, real-time analytical databases, streaming computation, operational visualisation and metrics…
More in Data, AI, Simulation & Analytics
All Data, AI, Simulation & Analytics programmes → · How to apply · Fees & financing · Scholarships