Data Scientist (Product Embedded ML)

Responsibilities

• Clean, structure, and analyse odour/sensor datasets for training and evaluation.
• Develop and optimise lightweight ML models (e.g. TensorFlow Lite) for embedded deployment.
• Collaborate with firmware engineers to integrate models into our smart trap prototypes (via platforms like Edge Impulse or custom firmware).
• Test model performance across lab, semi-field, and real-world settings; tune for reliability and accuracy.
• Maintain and document training pipelines, feature engineering methods, and model validation results.
• Support future product rollouts by adapting the detection model for different species, applications, and use environments.
• Assist in developing model-driven features for the platform (e.g. confidence scoring, anomaly detection, retraining logic).
• 5+ years of experience in applied data science or ML engineering roles.
• Strong applied experience in machine learning (ideally classification-focused).
• Proficiency in Python and relevant libraries (scikit-learn, TensorFlow, pandas, etc.).
• Experience working with real-world sensor or time-series data.
• Familiarity with embedded ML tools (e.g. TensorFlow Lite, Edge Impulse) or willingness to learn.
• Hands-on mindset — comfortable working in close collaboration with hardware, firmware, and product teams.
• Clear communicator who can explain model behaviour and constraints to non-technical collaborators.
• Experience in signal processing, low-power sensing, or IoT.
• Previous work on edge-deployable models.
• Interest or background in biology, chemistry, or environmental sensing.
• Familiarity with HDF5, metadata tagging, or sensor calibration pipelines.
• Experience in versioning datasets and ML workflows (e.g. MLFlow, DVC).
• Models in the field, not just the lab. You've taken at least one detection model from training through to embedded deployment on a live device — validated across lab, semi-field, and real-world conditions — with documented accuracy and reliability benchmarks that the product and R&D teams trust.
• A repeatable ML pipeline that others can build on. Training, feature engineering, validation, and versioning are no longer ad hoc processes. Working in collaboration with the Lead Data Scientist, you've established clean, documented workflows that mean the team isn't starting from scratch each time a new dataset, species, or use case arrives.
• The platform thesis is technically credible. You're actively contributing to dot.'s cross-domain ambitions — helping demonstrate that models and data structures built for one application (e.g. bed bug detection) can be adapted and transferred to others, supporting the Level 2 and Level 3 data value story we're building toward Series A.
• Firmware and hardware teams see you as a genuine collaborator. You're the bridge between data science and embedded systems — fluent enough in the constraints of the device environment that integration doesn't become a bottleneck, and trusted by engineers as someone who understands the real-world limits of what a model needs to do.
• dot.core has a stronger data foundation. Your work on sensor datasets - cleaning, structuring, metadata tagging, calibration pipelines - means dot.'s core data asset is more valuable, better labelled, and increasingly defensible as a proprietary dataset, not just a byproduct of discovery projects.
• Be part of a fast-growing, mission-driven company working at the intersection of science and technology.
• Your work will live in physical products that create measurable impact across public health, agriculture, and more.
• Collaborate with a multidisciplinary team of scientists, engineers, and domain experts.
• Work in an environment that values autonomy, innovation, and meaningful problem-solving.
• This role requires a minimum of 3 days per week in our Edinburgh office. Are you able to commit to this?
• Have you deployed a machine learning model onto an embedded or edge device (e.g. TensorFlow Lite, Edge Impulse, or custom firmware)?
• Have you worked with real-world sensor or time-series data, rather than only tabular, image, or simulated datasets?
• What is your current notice period, in weeks?
• Briefly describe a time you adapted a model or pipeline built for one application to a different use case or domain.
• A Cover Letter is required for this position, please confirm this has been included in your application along with your CV.
• Data science or ML engineering: 5 years (preferred)
• United Kingdom (required)