Pieter Cawood | Software & ML Engineer

Hi, I'm Pieter Cawood.

A

Machine learning, software, hardware and R&D management with extensive experience in industrial automation, developing software and product solutions for the 4th Industrial Revolution.

About

I am a machine learning and software engineer with a diverse background and extensive experience in industrial automation. With a passion for developing software solutions for the 4th Industrial Revolution, I embrace data-driven, edge, and cloud-based techniques to tackle challenging problems.

At Infinite Process Solutions, I serve as Lead R&D Software Engineer, where I architect MLOps and AI-based predictive maintenance solutions. My work includes managing the design team of IoT hardware, designing cloud infrastructure, building data pipelines with Python and Airflow, and deploying deep learning models with fast.ai and MLFlow.

Over the years, I have developed a deep understanding of various software standards in industrial automation. My experience spans leading and guiding teams through complex projects, including the development, offline programming, site commissioning, and support of over 100 discrete and batch manufacturing lines.

Languages: Python, .NET, SQL, JavaScript, TypeScript, PLC Programming
ML & AI: TensorFlow, PyTorch, fast.ai, Scikit-learn, XGBoost, LightGBM, CatBoost, MLFlow
Cloud & DevOps: Azure ML, Azure Databricks, Azure Data Factory, Docker, CI/CD, Airflow
WebApps: HTML, Django, Node.js, React.js, Next.js, nginx
Databases: SQL, NoSQL
Industrial Automation: Siemens, Rockwell, Beckhoff, CODESYS, Ignition

My academic journey includes an MSc in Artificial Intelligence (with distinction) and BTech in Electrical Engineering (top of class). I'm a Microsoft Certified Azure Data Scientist and have authored three published papers on state-of-the-art time-series forecasting.

Experience

Infinite Process Solutions

Lead R&D Software Engineer

Lead development of company products (hardware + software) including predictive maintenance solutions, remote SCADA software, and inventory management systems.
Architect of MLOps & AI-based predictive maintenance modeling. Designed and deployed cloud infrastructure, data pipelines with Python & Airflow, and deep learning models with fast.ai & MLFlow.
Designed IoT hardware architecture to interface with cloud-based predictive maintenance solution.
WebApp development for company products using modern full-stack technologies.
Provide Siemens PLC and HMI maintenance and support for clients including WaterNet and Zeiss (3rd level software support - Germany).
Tools: Python, Azure ML, Airflow, fast.ai, MLFlow, Django, Docker, Siemens PLC/HMI

2023 - Present

Pliant

Software Engineer

Developed C# desktop applications for HMI systems using MVC-WPF pattern with Entity Framework for database logging and statistics.
Designed Power BI reports for fruit quality analysis, processing vast data sources for the Rijk Zwaan Paprika Fruit Quality Grading project.
Integrated AI models into production systems: implemented C++ DLL inference to detect defects using YOLO v6 trained models.
Modified PLC communication systems using TwinCAT ADS .NET and performed MVTec Halcon vision scripting.
Conducted data-driven research and proof-of-concept studies for new client projects.
Tools: C#, .NET, WPF, Entity Framework, TwinCAT ADS, Power BI, YOLOv6, MVTec Halcon

2022 | The Netherlands

Tomamate (Self-employed)

Software Controls Engineer

Designed Ignition HMI systems for multiple battery production lines at GM T1 Flex (Michigan), GM Oshawa, and DWFritz Battery Lines (Oregon).
Led PLC programming projects including architectural design, offline programming, and on-site commissioning for automotive and battery production facilities.
Created and improved control system software architectures, ensuring optimized functionality and resilient designs.
Led teams to roll out software updates and managed large-scale hardware setup, commissioning, and troubleshooting.
Ensured customer satisfaction during Factory Acceptance Test (FAT) checks and implemented system safety hardware design updates.
Tools: Ignition HMI, PLC Programming, Industrial Automation Systems

2021 - 2022 | USA and Canada

DES Group

Software Controls Engineer

Delivered large-scale industrial automation solutions across South Africa, Thailand, Mexico, and the USA for global automotive clients including Ford, BMW, VW, Magnetto, and Tenneco.
Key projects executed:
- Ford P375 – FTM Thailand
- Ford P375 Upgrades – FTM Thailand
- Ford P375 – MAP Michigan, USA
- BMW G01 – Magnetto South Africa
- BMW G20 & G42 – SLP Mexico
- Tenneco ACL – DES Group SA Warehouse
PLC programming including architectural design, offline development, on-site commissioning, and production support.
Setup, commissioning, and optimization of Human-Machine Interfaces (HMIs).
Designed and enhanced control system software architectures with robust sequencing logic, alarming, and diagnostics.
Performed virtual commissioning using Process Simulate scripts, simulating hardware behavior over OPC-UA.
Developed C# utilities to support PLC code development and automation workflows.
Integrated robotics, vision systems, high-precision servo motion control, and real-time sensor feedback.
Designed OT networking architectures with full MES and SCADA integration for coordinated production management and data-driven decision-making.
Technologies: PLCs, HMIs, Robotics, Servo Motion Control, OPC-UA, C#, Vision Systems, MES, SCADA

2015 – 2020 | South Africa, Thailand, Mexico & USA

Some Projects

                
                  VersaNode
                
                  Modular smart node for industrial automation and edge computing (Raspberry Pi 5 Compute host).
                
                  Accomplishments
                Tools: Embedded Linux, Linux, React.js, JavaScript, Product Development
Led product vision and technical roadmap as Product Owner for a next-gen edge node.
Defined modular service packages: PLC runtime, edge gateway, data acquisition, MQTT/OPC UA broker, firewall, VPN, database services.
Designed architecture for software-addressed sub-modules (NVMe, GSM, Edge TPU, high-speed analog/digital I/O) via backplane bus.
Delivered interoperability across OT/IT systems with cross-discipline stakeholder coordination and release planning.
Connectivity: dual gigabit LAN, WiFi 7, eSIM cellular, and local/remote I/O expansion.

                  DeepPID
                
                  ML-augmented PID controller suite (PyTorch) with real-time simulation + GUI benchmarking.
                
                  Accomplishments
                Tools: Python, PyTorch, Tkinter, Matplotlib, MPC, PID Control
Bridged classical control (IMC, Cascade PID) with neural adaptive controllers (GRU, MLP, Transformer) and a hybrid MPC variant.
Built a modular “controller zoo” + extensible interfaces for rapid integration of new controllers/plants.
Implemented multiple dynamic plant simulations (first-order, back-pressure coupled, two-tank, quadcopter) for comparative benchmarking.
GRU-based controller learns online from recent history to outperform fixed-gain PID under nonlinear/drifting dynamics.
Delivered real-time evaluation via a Tkinter + Matplotlib interactive GUI.

                  BrowserRL
                
                  Rainbow DQN agent trained via Playwright automation to play a real browser game.
                
                  Accomplishments
                Tools: Python, Reinforcement Learning, Playwright, OpenCV, Deep Q-Learning
Integrated browser state extraction with a custom reward function for iterative training episodes.
Built a pipeline combining web automation, real-time perception, and RL-based decision-making.
Demonstrated RL interaction with dynamic browser environments (gameplay + optimization loop).
Planned extensions: multi-game generalization and improved policy transfer across similar environments.

                  PyAsyncBTrack
                
                  High-performance Asynchronous Backtracking (ABT) solver for DCSPs with modern heuristics.
                
                  Accomplishments
                Tools: Python, Algorithms, Multi-agent Systems, AI
Implemented ABT as a centralized solver framework for distributed constraint satisfaction problems (DCSPs).
Added MRV/LCV heuristics, conflict-directed backjumping with nogood learning, and optional AC-3 pre-pruning.
Supported domain reshuffling restarts for search diversity and multi-solution enumeration.
Delivered clean, fully-typed APIs optimized for asynchronous execution and integration in AI pipelines.

                  M-TA-Prioritized
                
                  Modified TA-Prioritized (AAAI 2019) implementation for multi-agent pickup & delivery with improved deadlock avoidance.
                
                  Accomplishments
                Tools: Python, Multi-agent Systems, AI
Implemented TA-Prioritized for MAPD in large-scale dynamic environments.
Extended baseline with a modified deadlock avoidance mechanism to reduce congestion.
Improved makespans in dense task scenarios, focusing on scalability and robustness.
Emphasized distributed decision-making and reliable coordination under high task density.

                  FFORMA-ESRNN
                
                  Forecasting framework extending FFORMA with ESRNN + hybrid ensemble stochastic neural networks (ESNNs).
                
                  Accomplishments
                Tools: Python, Time Series, Deep Learning, Ensemble Learning, Statistics
Extended FFORMA ensemble selection with learning from ESRNN and additional hybrid models.
Enhanced meta-learning architecture using ensemble stochastic neural networks (ESNNs).
Improved generalization and robustness across diverse forecasting tasks and datasets.
Contributed to automated forecasting pipelines with interpretable hybrid deep learning systems.