Capability Engineer – Next-Gen Pet Center Technologies

Job Description:
The Capability Engineer designs, builds and scales the next-generation technology backbone of Royal Canin’s R&D Pet Centers – cameras and smart sensors, edge devices, local infrastructure of GPU servers and on-premise machine-learning models – so that every meal and every protocol generates reliable, decision-grade behavioural and intake data. Based in Lewisburg (RLE), the role pilots and proves these technologies at RLE, then leads their standardised scale-up to Aimargues (RCA) and, over time, to the wider Mars Pet Center network. It takes AI-assisted observation (palatability, digestibility, welfare and food-safety/quality signals) from prototype to secure, cross-site operation, releasing associates from manual scoring and feeding harmonised datasets into Monarch and then feed Royal Canin’s multi-modal predictive models. Working with Pet Center operations at both sites, R&D, Digital/IT, MGS Services, MGS Labs and external partners, the engineer makes RLE the technology pilot site and ensures that what works there is replicated at RCA, accelerating learning cycles and speed to market.
What are we looking for?
Bachelor’s or Master’s degree in computer or electrical engineering, computer science, data science, mechatronics or a related discipline, or equivalent practical experience. Training or certification in networking, server/infrastructure, edge or cloud computing, cybersecurity, AI/machine learning or project management is preferred.
Hands-on experience designing and deploying camera/video systems and IoT sensors in operational environments (IP cameras, PoE networks, video management, time synchronisation, lighting and placement).
Practical experience putting computer-vision/machine-learning models into production: Python, a deep-learning framework (e.g. PyTorch), containerisation (e.g. Docker), model optimisation for GPU/edge (e.g. ONNX, TensorRT) and validation using precision/recall-type metrics; familiarity with edge devices (e.g. NVIDIA Jetson, Raspberry Pi) is an advantage.
Experience verifying and monitoring ML models in production: validation on held-out subjects and continuous data, drift detection, model and dataset versioning, retraining workflows and documentation of model performance over time.
Rapid prototyping mindset: builds working prototypes quickly, including with AI-assisted coding tools, and turns successful ones into maintainable, documented, scalable solutions; experience integrating or customising ML tooling (annotation, training, experiment tracking) is an advantage.
Working knowledge of server, storage and network infrastructure (Linux/Windows, GPU servers, virtualisation, backup, monitoring, TCP/IP, identity and access, firewalls and isolated/air-gapped designs) and cybersecurity fundamentals.
Data engineering skills: data pipelines, metadata models, data-quality checks and integration with data lakes or research data platforms.
Experience taking technology from pilot to multi-site scale: sizing, business cases, standardisation, vendor management, documentation, change management and user training; comfortable working across sites, cultures and time zones (US/Europe).
Strong team player in cross-functional project teams: translates between scientists, animal-care operators, facilities and IT; builds trust with end users, troubleshoots complex issues, analyses root causes and communicates clearly with technical and non-technical stakeholders.
Experience in an R&D, laboratory, animal-care, manufacturing or similarly regulated environment is preferred; exposure to animal-behaviour data or annotation/labelling tools is an advantage.
What will be your key responsibilities?
Own the cross-site Next-Gen Pet Center effectiveness technology roadmap: translate R&D and operational needs (palatability/deference, digestibility, food safety/quality, welfare) at RLE and RCA into one phased technology architecture and deployment plan – pilot at RLE, scale to RCA; own each technology across its lifecycle – requirements, selection, implementation, validation, documentation and handover for scale-up, support upgrades, continuity and retirement.
Pilot at RLE, then scale the camera and sensor network: specify, source, install and maintain cameras, connected scales, smart litter boxes, environmental and other IoT sensors; prove a first fully instrumented prototype room at RLE, extend to site-wide coverage (feeding stations, panel rooms, lodges & gardens), then replicate the validated set-up at RCA, adapted to local rooms, animals and constraints, with consistent placement, lighting, resolution and camera-zone calibration so models perform reliably at both sites.
Build and run the local compute platform: size, stand up and operate edge devices (e.g. NVIDIA Jetson-class), on-premise GPU servers, storage, backup and networking – first at RLE, then as a standard, replicable configuration at RCA; start with an air-gapped environment for highly confidential data and move to governed connectivity (Mars-approved networking, high-performance computing) as approvals allow. Size infrastructure by separating model training/retraining from production inference, and real-time from batch processing.
Industrialise local machine-learning models: deploy, containerise and run computer-vision pipelines (e.g. object detection, segmentation/tracking and video-classification models) on local infrastructure, using standard ML frameworks and inference-optimisation tools (e.g. PyTorch, ONNX, TensorRT); validate them first on the RLE animal population and then at RCA, and fine-tune them for new breeds, rooms, sites and use cases with R&D behaviour scientists and model developers.
Run the ML model verification programme so performance does not slip: own a verification and validation plan for every model in use – acceptance criteria agreed with R&D before go-live (e.g. precision, recall, F1, false-positive rate by behaviour), independent testing on continuous video from the target room and site, and periodic re-verification against expert-coded samples from the ’expert validates’ step. Monitor performance and data drift as the pet population changes (animal turnover, new breeds, ages, sizes, coat colours) and as rooms, feeders, cameras, lighting, diets and protocols change; define triggers for recalibration, retraining or rollback; version-control models, training data and thresholds under change control so every result is traceable to the model version that produced it; and re-verify before any roll-out to a new room or to RCA.
Integrate and adapt shared ML tooling: integrate, configure and, where needed, modify the shared tools that the ML lifecycle depends on – annotation and labelling, dataset management, model training, experiment tracking and model registries (e.g. open-source annotation platforms, MLflow-type tools) – so researchers, model developers and Pet Center teams at both sites label, retrain and validate models on one common, governed toolset.
Prototype fast, then scale what works: write and AI-assisted-code local prototypes – apps, scripts, dashboards, data-capture tools and model pipelines (e.g. using GitHub Copilot-type coding assistants) – test them with Pet Center users, and once proven successful harden them into documented, supportable solutions, to scale them into facility operations at RLE and RCA in collaboration with VCO IT, MGS and DT teams, and link them to Monarch.
Embed AI into Pet Center workflows: implement the ‘AI pre-codes, expert validates’ workflow (batch processing followed by expert review and correction in shared annotation tools), move associates from paper to tablet-based capture, and integrate new data-capture steps into palatability, digestibility and research protocols at both sites in line with the Every Meal Matters philosophy – without compromising animal welfare or study integrity.
Work as one team with Pet Center, Facilities and R&D colleagues: co-define use cases, requirements and success criteria with Pet Center teams at RLE and RCA (managers, team leaders, animal-care, palatability/digestibility and welfare technicians, veterinarians) and involve them early in prototyping, testing and roll-out; partner with R&D Facilities colleagues (Methodology Expert, Method Engineer, operations and data technicians, pilot plant and site maintenance/operations) on installation, infrastructure, method integration and shared standards; work with R&D scientists and project teams (palatability, digestibility, behaviour and welfare, nutrition,