TinyML LLM Optimization Engineer
Budget: €1,500 – €3,000 EUR
I’m looking for a TinyML / Edge AI specialist who can squeeze Large Language Model capabilities into very small neural networks that run on constrained hardware.
Scope of work
• Collect and curate compact yet representative datasets.
• Design and run a full slate of experiments focused on pruning, quantization, and knowledge distillation.
• Keep clear experiment logs and versioned notebooks so every tweak is reproducible.
• Generate benchmark results that show latency, memory footprint, and accuracy trade-offs on target devices (MCUs,chips, or other edge platforms).
• Interpret those results, recommend next moves, and iterate until we hit the sweet spot between model size and performance.
What I expect
– Thoughtful selection of toolchains such as TensorFlow Lite Micro, ONNX, or PyTorch Mobile—whatever best fits the target.
– Clean, well-documented code and scripts ready for handoff or direct deployment.
– A concise final report summarizing datasets, experiment logs, and benchmarks, plus annotated source files.
If you love making big models tiny and efficient, let’s talk.
Scope of work
• Collect and curate compact yet representative datasets.
• Design and run a full slate of experiments focused on pruning, quantization, and knowledge distillation.
• Keep clear experiment logs and versioned notebooks so every tweak is reproducible.
• Generate benchmark results that show latency, memory footprint, and accuracy trade-offs on target devices (MCUs,chips, or other edge platforms).
• Interpret those results, recommend next moves, and iterate until we hit the sweet spot between model size and performance.
What I expect
– Thoughtful selection of toolchains such as TensorFlow Lite Micro, ONNX, or PyTorch Mobile—whatever best fits the target.
– Clean, well-documented code and scripts ready for handoff or direct deployment.
– A concise final report summarizing datasets, experiment logs, and benchmarks, plus annotated source files.
If you love making big models tiny and efficient, let’s talk.