Diploma in On-Device Runtimes (TFLite/ONNX/TensorRT)
Online postgraduate diploma of 800 hours in On-Device Runtimes (TFLite/ONNX/TensorRT): 8 modules covering on-device inference, LiteRT and TensorFlow Lite, ONNX Runtime, NVIDIA TensorRT on Jetson, alternative compilers and runtimes, inference performance optimisation and product integration, 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
Select suitable formats and runtimes for on-device inference.
Convert and deploy models with LiteRT and TensorFlow Lite.
Optimise ONNX models with different execution providers.
Build optimised TensorRT engines on Jetson platforms.
Compile models with TVM, ExecuTorch and other runtimes.
Optimise inference latency, memory and power consumption.
Integrate models into products with validation and monitoring.
Who it's for
- Embedded AI engineers deploying neural networks on edge hardware.
- Model deployment engineers moving models from training to production devices.
- Vision application developers building real-time detection and classification.
- Inference performance engineers chasing latency, memory and energy targets.
- Mobile software engineers adding on-device AI to Android and iOS apps.
- Edge platform architects choosing runtimes and accelerators.
- Engineering or science graduates looking to enter the consumer electronics sector.
What you'll work with
- ONNX Specification – LF AI & Data (open neural network exchange format)
- Khronos NNEF 1.0 (Neural Network Exchange Format)
- ISO/IEC 23053:2022 (framework for AI systems using machine learning)
- IEEE 754-2019 (floating-point arithmetic)
- ISO/PAS 8800:2024 (road vehicles: safety and artificial intelligence)
- Regulation (EU) 2024/1689 – AI Act (harmonised rules on artificial intelligence)
- ISO/IEC 42001:2023 (AI management systems)
- LiteRT (TensorFlow Lite)
- ONNX Runtime
- NVIDIA TensorRT
- Apache TVM
- ExecuTorch
8 modules
1.1Advantages of local inference: latency and privacy1.2Export cycle from training1.3Computational graphs and operators1.4Formats: TFLite, ONNX, TorchScript and GGUF1.5Hardware: CPU, GPU, NPU and DSP1.6Delegates and execution providers1.7Memory and power constraints1.8Runtime comparison and selection criteria1.9Metrics: latency, throughput and power consumption1.10Lab: exporting a PyTorch model to ONNX
2.1Conversion with TFLiteConverter2.2FlatBuffers and model structure2.3GPU, NNAPI and XNNPACK delegates2.4Quantisation integrated into conversion2.5Custom operators and kernels2.6TensorFlow Lite Micro for microcontrollers2.7Integration on Android and embedded Linux2.8Benchmark tool and profiling2.9Memory management and the tensor arena2.10Practical: classifier on Raspberry Pi with XNNPACK
3.1Opsets and format versioning3.2Export with torch.onnx and dynamo3.3Graph optimisation and fusions3.4Execution providers: CUDA, TensorRT and OpenVINO3.5Static and dynamic quantisation in ONNX3.6ONNX Runtime Mobile and reduced builds3.7C++, Python and C# APIs3.8Debugging with Netron and onnx-checker3.9Dynamic shapes and symbolic dimensions3.10Lab: cross-platform deployment with ONNX Runtime
4.1Engine building and plan files4.2FP16, INT8 and FP8 precision4.3INT8 calibration with calibrators4.4Plugins and custom layers4.5Optimisation profiles with dynamic shapes4.6DLA and GPU on Jetson Orin4.7DeepStream for video pipelines4.8trtexec and Nsight Systems4.9Triton Inference Server at the edge4.10Workshop: object detector with TensorRT on Jetson
5.1Apache TVM and autotuning5.2ExecuTorch for PyTorch on devices5.3Core ML in the Apple ecosystem5.4OpenVINO on Intel platforms5.5Vendor NPU runtimes5.6MLIR and IREE5.7llama.cpp and local language models5.8WebGPU and ONNX Runtime Web5.9Portability comparison5.10Practical: compiling with TVM for ARM
6.1Per-operator profiling and bottlenecks6.2Operator fusion and memory layouts6.3Multithreading, affinity and scheduling6.4GPU-accelerated preprocessing6.5Batching and pipelining6.6Thermal management and throttling6.7Energy consumption per inference6.8Start-up time and model loading6.9Peak memory optimisation6.10Lab: optimised end-to-end latency
7.1Integration into C++ and Android applications7.2Updating models in the field7.3On-device accuracy validation7.4Numerical regression testing7.5Runtime version management7.6Model security and protection7.7Compliance with the EU AI Act7.8Functional safety and ISO/PAS 88007.9Telemetry and monitoring7.10Case study: vision assistant for an industrial camera
8.1Selection of the model and platform8.2Model export and conversion8.3Deployment on two runtimes8.4Optimisation and quantisation8.5Comparative benchmark between runtimes8.6Integration into an application8.7Validation and testing8.8Recommendations report8.9Report 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 specification and applied exercises. The working software is LiteRT (TensorFlow Lite), ONNX Runtime, NVIDIA TensorRT, Apache TVM and ExecuTorch, in free, academic or evaluation versions. Testing is studied through real procedures: latency benchmarks on Jetson and Raspberry Pi, model conversion between formats and per-operator inference profiling. Assessment is continuous, module by module, plus the final project.
Standards covered: ONNX Specification (LF AI & Data), Khronos NNEF 1.0, ISO/IEC 23053:2022, IEEE 754-2019, ISO/PAS 8800:2024, Regulation (EU) 2024/1689 (AI Act) 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 LiteRT (TensorFlow Lite), ONNX Runtime, NVIDIA TensorRT, Apache TVM and ExecuTorch, 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
Postgraduate Diploma in Railway RAMS and Rolling Stock
Online 800-hour Postgraduate Diploma in Railway RAMS and Rolling Stock. 8 modules covering EN 50126-1:2017, EN 50126-2:2017 and Isograph Reliability Workbench. Assigned tutor and enrolment open.
Postgraduate Diploma in ISO 13485, QMS and Validation
Online Postgraduate Diploma of 800 h in ISO 13485, QMS and Validation. 8 modules covering ISO 13485:2016, Regulation (EU) 2017/745 and Greenlight Guru. Assigned tutor and…
Postgraduate Diploma in UDS/DoIP Diagnostics and Vehicle Security
Online 800 h Postgraduate Diploma in UDS/DoIP Diagnostics and Vehicle Security. 8 modules covering ISO 14229-1:2020 (UDS), ISO 13400-2:2019 (DoIP) and Vector CANoe and CANalyzer. Assigned…
More in Electronics, Embedded Systems & Telecom
All Electronics, Embedded Systems & Telecom programmes → · How to apply · Fees & financing · Scholarships