Ian Franson Mendonca

Developed a web-based GUI for segmenting and annotating cracks in power electronics using JavaScript and Flask, optimizing data workflows for machine learning. Designed an SQL-based relational database for efficient data management. Created and optimized deep learning models for defect detection with semantic segmentation and classification, achieving 80% accuracy. Implemented Autoencoders, PCA, and clustering techniques. Containerized the model with Docker and deployed it.

I developed a Retrieval-Augmented Generation (RAG) system using LangChain and LlamaIndex for processing large text volumes, and Pinecone for fast semantic searches. This enabled the use of LLMs like GPT-4 for generating dynamic responses. Additionally, I built an NLP system for sentiment analysis and spam detection based on Random Forest algorithms.

For my Master’s thesis at Bosch, I developed ML algorithms to optimize thermal paste dispensing in power electronics production. I designed data pipelines using SQL and applied PCA for data reduction. I also developed an AutoML pipeline with Optuna, optimizing parameters using gradient-based methods. The model was deployed on Databricks, achieving 80% coverage efficiency, and enhancing production quality and performance in power modules. Skills: Python, Tensorflow-Keras, AutoML, Scikit-Learn, Matplotlib,

Achievements: •Predictive Maintenance for Industrial Equipment: Collaborated with the manufacturing dept to develop a predictive maintenance system. We could forecast equipment failures by analyzing real-time sensor data for spindles, enabling proactive maintenance scheduling, and reducing downtime and maintenance by 20%. •Led the implementation of SAP S4/HANA in the organization, streamlining business and improving data management resulting in a 25% increase in efficiency for design and assembly personnel.