Nearhuman (Scootrr)
Senior Machine Learning Engineer – Edge AI
How your CV stacks up
Upload your CV to see how well it fits this job role
?%
About Nearhuman
Nearhuman develops privacy-first edge-AI systems that allow advanced perception models to operate directly on low-power devices. We work with computer vision, embedded systems and multimodal sensor data to build technology for mobility, robotics and other environments where power, memory, latency and connectivity are constrained.
We are looking for an experienced machine learning engineer to strengthen our internal technical team and turn advanced research methods into robust, reproducible and deployable systems.
The role
As Senior Machine Learning Engineer – Edge AI, you will take engineering ownership of the implementation, evaluation and deployment of Nearhuman’s machine-learning models.
You will work closely with our research leadership, engineering team and external research collaborators. Your responsibility will be to translate technical concepts and research methods into well-engineered training pipelines, reproducible experiments and models that can operate reliably on low-power hardware.
This is not a purely academic role. You will be expected to write production-quality code, work with imperfect real-world data, profile models on physical devices and support integration into commercial systems.
The scientific direction of the company’s research will be owned by the research leadership. You will be responsible for ensuring that research ideas become functioning, testable and deployable technology.
Key responsibilities
Model development and experimentation
- Implement efficient neural-network architectures using PyTorch.
- Translate research papers and mathematical methods into reliable software.
- Develop and compare full-precision, quantised and low-bit models.
- Build models combining visual and time-series sensor data.
- Investigate optimisation instability, modality imbalance and model degradation.
- Run controlled experiments across multiple configurations and random seeds.
- Analyse failures and recommend practical engineering improvements.
Training and evaluation
- Build reproducible machine-learning training and evaluation pipelines.
- Manage dataset splits, preprocessing, augmentation and experiment configurations.
- Develop automated benchmarking across model architectures.
- Measure accuracy, precision, recall, F1, calibration, latency, memory use and energy consumption.
- Maintain experiment tracking, model versioning and reproducibility records.
- Ensure results can be independently reproduced by other engineers.
- Clearly distinguish proven results from assumptions and early-stage findings.
Multimodal data engineering
- Build pipelines for combining visual and time-series sensor data.
- Develop data-cleaning, timestamp-alignment and labelling processes.
- Work with public research datasets and internally collected data.
- Identify data-quality problems, missing information and labelling inconsistencies.
- Ensure data is handled according to licensing, confidentiality and data-protection requirements.
Edge deployment and optimisation
- Deploy and profile models on ARM-class processors and embedded Linux systems.
- Optimise models for latency, memory, energy use and thermal constraints.
- Convert and deploy models using ONNX, TFLite, TVM, ExecuTorch, OpenVINO or similar tools.
- Investigate CPU, NPU and embedded-accelerator deployment.
- Support integration into embedded camera and sensor-processing systems.
- Develop repeatable device-level benchmarking and power-measurement procedures.
- Diagnose performance bottlenecks across software and hardware.
- Validate models using laboratory testing and real-world data.
Reasons to use Rodeo
Start with a chat, not a search bar
Grad scheme, placement, apprenticeship? Not sure what you want yet — that's fine. Your agent talks it through with you and turns "I have no idea" into a shortlist.
It searches the market for you
Every day your agent scans the market matching roles against what actually matters to you, not just keywords on a CV.
Only hits
No noise. No "maybe this fits." Just roles with a clear explanation of why they're right — and where to focus when applying.
Software engineering
- Write clear, modular and maintainable Python code.
- Contribute to C or C++ components where required.
- Maintain Git repositories, code reviews, testing and documentation.
- Build containerised and reproducible development environments.
- Develop automated tests for data pipelines and model components.
- Maintain traceability between datasets, code versions, model weights and results.
- Improve the quality and reliability of Nearhuman’s machine-learning processes.
Collaboration and documentation
- Work with research leadership to translate scientific objectives into engineering plans.
- Provide evidence on feasibility, implementation complexity and deployment constraints.
- Participate in technical design reviews and challenge assumptions constructively.
- Document implementation decisions and experimental outcomes.
- Produce reproducibility packages that allow Nearhuman to retain technical capability internally.
- Maintain accurate records of experiments and technical contributions.
- Prepare architecture documents, benchmark reports and deployment guides.
- Identify potentially novel implementation or optimisation outcomes.
Commercial and partner support
- Support technical discussions with customers and technology partners.
- Translate requirements into measurable technical acceptance criteria.
- Prepare technical demonstrations and diligence materials.
- Explain model limitations, deployment requirements and benchmark results clearly.
- Support integration planning across mobility, robotics, autonomy and industrial applications.
Essential experience
- Strong practical experience developing and training neural networks using PyTorch.
- Experience in computer vision, multimodal learning, sensor fusion, robotics perception or a related area.
- Strong Python and software-engineering skills.
- Experience implementing techniques from research papers.
- Experience designing and running controlled machine-learning experiments.
- Understanding of model evaluation, calibration, class imbalance and statistical variation.
- Experience with quantisation, low-bit inference, pruning, knowledge distillation or model compression.
- Experience building reproducible training and benchmarking pipelines.
- Experience using Git, automated testing and containerised environments.
- Experience deploying or profiling models on edge devices or embedded Linux systems.
- Ability to diagnose model-performance and data-quality problems independently.
- Ability to document technical work and communicate problems early.
- A relevant degree or equivalent professional experience.
Get help with your application
Your very own career expert that helps elevate your application to the next level.
Desirable experience
- Binary neural networks or native low-bit training.
- Multimodal, vision-language or embodied-AI models.
- Visual data combined with IMU, GPS, CAN or other time-series sensors.
- Model calibration and uncertainty measurement.
- Energy and latency profiling on physical devices.
- ARM processors, NPUs or embedded accelerators.
- ONNX, TFLite, TVM, ExecuTorch or OpenVINO.
- C or C++ development.
- NVIDIA Jetson, Raspberry Pi or comparable edge platforms.
- Real-time camera pipelines, GStreamer or RTSP.
- MLflow, Weights & Biases or similar experiment-tracking tools.
- Publications, patents or open-source work in efficient AI or computer vision.
- Experience working in a startup or small technical team.
- Experience in robotics, mobility, autonomous systems or industrial technology.
What success looks like in the first three months
You will be expected to:
- Understand Nearhuman’s hardware, software, datasets and model environment.
- Establish a clean and reproducible machine-learning workflow.
- Implement or verify an initial full-precision baseline.
- Develop and compare optimised or low-bit model variants.
- Build an initial multimodal data-processing pipeline.
- Establish benchmarks for accuracy, calibration, latency, memory and energy.
- Run repeatable experiments across multiple configurations.
- Deploy at least one model on an edge-compute platform.
- Document the architecture, datasets, experiments and deployment process.
- Produce a reproducibility package that another engineer can run.
- Recommend the next stage of model development and deployment.
Personal qualities
We are looking for someone who:
- Takes ownership and follows work through to completion.
- Is comfortable working in an early-stage company.
- Communicates blockers, uncertainty and failed experiments openly.
- Values reproducibility, testing and documentation.
- Can distinguish promising experiments from commercially usable evidence.
- Enjoys working across research, software, embedded hardware and real-world testing.
- Can work closely with senior researchers without needing to own the scientific agenda.
Why join Nearhuman
- Build AI systems operating under genuine power, memory and latency constraints.
- Work with real-world camera and multimodal sensor data.
- Collaborate with experienced researchers, engineers and industry partners.
- Help shape the engineering foundations of a growing deep-technology company.
- See your work move from experiments into physical embedded systems and commercial deployments.
How to apply
Please submit:
- Your CV.
- A short description of the most technically difficult machine-learning system you have built.
- An example of a model you have trained, optimised or deployed on constrained hardware.
- Links to relevant repositories, publications or technical projects.
- Your UK right-to-work status, availability and expected salary range.
“It took my CV and asked me questions relevant to understanding what kind of jobs to suggest for me. Suggestions were almost perfect. Jobs were exactly what I’ve been looking for.”
Jessica, London
Skills
Location