Zakaria Rada

Building UnixonAI Edge Platform, an enterprise edge AI infrastructure that enables organizations to run generative AI securely within their own environment — with full control over data, governance, and auditability. • Developing on-prem AI platform for secure internal LLM and document intelligence • Designing cluster-to-node runtime architecture for private AI deployment • Building governance layer: access control, policy enforcement, and AI audit logging

As Team Lead for AI Software Engineering on WoRV (World model for Robotics and Vehicle control), I led the design and implementation of the robotic software system and backend infrastructure for Korea’s first physical AI deployment in autonomous agricultural machinery. Key Contributions • Robot Control & Hardware Integration: Led development of the robotic software system with GPS, CAN bus, IMU sensors, and actuator control; contributed to CAN protocol design and hardware system validation. • Backend Infrastructure: Designed scalable cloud architecture using AWS (EKS, RDS, DynamoDB), Docker, and Python (FastAPI, Django) for real-time task orchestration across 100+ robots. • Real-Time Communication: Engineered low-latency pipelines via WebRTC SFU, Redis, and RabbitMQ to support live telemetry and operator control. • AI & MLOps: Integrated vision-based models for path planning and obstacle avoidance; deployed MLOps pipelines for continuous AI model delivery. • LLM-Powered Interface: Enabled natural language-based mission control using LLMs, allowing operators to give intuitive commands. • CI/CD Automation: Built GitLab CI/CD workflows for reliable, rapid deployment of microservices and AI features. • Commercialization: Spearheaded the successful deployment of WoRV in GINT's autonomous sprayer “PluvaSS,” delivering the first commercialized physical AI system in Korean agriculture.

Responsibilities: • Led AI strategy development, prioritizing innovation and scalability. • Built an MLOps platform on AWS, integrating CI/CD, Docker, and Kubernetes for cost-effective model iterations. • Implemented real-time model monitoring with Prometheus, Grafana, and Kibana, enhancing operational efficiency. • Optimized AI pipelines using multi-GPU setups and Nvidia Triton Server for high-throughput processing. • Launched Snowflake data warehousing, streamlining data access and analytics workflows. • Collaborated with DevOps and data teams to develop ETL pipelines, ensuring seamless AI integration into production. Key Projects: • Time Series Analysis & Price data: Led development of LSTM, ARIMA, and KNN models for price imputation and forecasting, increasing data accuracy and reducing reporting cycles. • NLP & Media Automation: Built NLP models (BERT, DistilBERT, NER) for media classification and product tagging, automating 90% of manual tasks and improving GPU performance with Hugging Face Transformers. • Trade Intelligence System: Developed a market opportunity detection system using GPT-4, Neo4j, and OpenSearch, improving query accuracy and decision-making. • Recommendation Systems & User Event Data: Designed recommendation engine using Sentiment Analysis and Collaborative Filtering, improving user engagement. Implemented event data tracking and TensorFlow to optimize feature engineering and accuracy. • Data Engineering: Built ETL pipelines to transfer data between Amplitude, Snowflake, and RDS, reducing integration times and enhancing real-time analytics. • R&D and LLM Model Development: Experimented with new LLM models (Llama, Mistral, etc.), benchmarking and fine-tuning them using LoRA and QLoRA for future applications.

Department: Lifting Division, R&D Role: AI - IoT Software Development Lead. Responsibilities: • Diagnostic Application Development: Led the development of BLE-based diagnostic apps for escalators and elevators on Android and iOS, using STM32 tools for PCB design and testing, ensuring real-time fault detection. • IoT and Bluetooth Integration: Integrated IoT sensors and BLE technology (ESP32 chips) into PCB boards for secure, real-time data transfer, improving device communication with custom encoding and decoding processes. • Predictive Maintenance & Machine Learning: Developed predictive maintenance algorithms (LSTM, Random Forest, SVM), improving fault detection and reducing downtime by 25%. Applied cross-validation and anomaly detection to enhance reliability. • Application Deployment: Launched and managed three mobile apps—EscalatorHealth, ElevatorInsight, and ElevatorRescue—leveraging BLE and machine learning for diagnostics, maintenance, and emergency response. Deployed ML models for real-time predictions. • Firmware & Data Management: Enabled remote firmware updates by transferring bin files from Android to PCB boards. Managed real-time data collection and secure storage in a centralized database. • Secure Data Transfer: Implemented Bluetooth encryption across apps to ensure secure data transfer between devices and PCB boards, preventing unauthorized access. Key Projects: • EscalatorHealth App: Developed BLE-based app for real-time escalator monitoring. Trained LSTM models on an 8-year dataset, reducing failures by 20% and cutting diagnostic time by 99%. • ElevatorInsight App: Created an elevator maintenance tool using SVM and Random Forest models, improving predictive maintenance and equipment reliability. • ElevatorRescue App: Built an emergency app using LSTM models to predict system failures, reducing response times by 99% and improving safety protocols.

Field: IoT, AI, Smart Farms, Vertical farming Systems Role: Led a team of software developers specializing in IoT Systems and Hardware Implementation. Responsibilities: • Led the development of AI systems for optimizing vertical farming operations. • Developed smart farming apps for real-time monitoring and control. • Integrated IoT devices with AI to automate and enhance farming processes. • Built a custom control panel for IoT devices, improving greenhouse management and installation. • Designed scalable systems with robust AI and database architectures. • Prototyped AI-driven products to increase system reliability and efficiency. Key Projects: • Vertical Farming System Design: Built a mini greenhouse using 3D modeling for IoT testing, increasing crop yield by 25% through real-time optimization. • Smart Farm App: Developed an AI-driven Android app for real-time farm monitoring and control, reducing manual intervention by 40%. • Custom IoT Control Panel: Designed a control panel for easy IoT device installation and management, improving efficiency and reducing setup time. • Greenhouse Dashboard: Created a web-based dashboard with AI analytics for real-time tracking, enhancing decision-making and crop management. • AI-Powered Crop Monitoring: Built a camera-based system using Mask R-CNN for crop health monitoring, detecting diseases with 80% accuracy and reducing crop loss by 30%. • Predictive Crop Yield Modeling: Developed LSTM models to forecast crop yields with 35% improved accuracy, enabling better resource planning. • Automated Disease Detection: Enhanced detection systems with AI (Mask R-CNN, YOLOv5) for early disease identification, reducing crop damage and improving health management.

Researcher Projects – Industrial & Systems Engineering Department Gyeongsang National University - Graduate School, Jinju, Korea

Thesis title: Development of Intelligent Alert System Based on Control Rules and IoT