Sergiu Dinu

Designed and deployed multimodal Al pipelines using PyTorch, TensorFlow, and Hugging Face Transformers, powering large-scale LLM and generative Al projects that supported global clients in automotive and life sciences sectors. Developed an intelligent data labeling automation system with Python, FastAPI, and Kubernetes, leveraging active learning to reduce manual annotation costs by 35% and improving annotation throughput for various enterprise projects. Implemented a synthetic data generation framework using Stable Diffusion, Whisper, and LLaMA fine-tuning, enabling clients to expand scarce datasets in speech recognition and computer vision, which improved downstream model accuracy. Built a scalable data quality validation suite with Apache Spark, Airflow, and MLflow, processing 60M+ multimodal samples, ensuring dataset integrity, and maintaining compliance with strict privacy standards. Delivered custom LLM fine-tuning and deployment solutions using LoRA, PEFT, and vector databases, enabling localized, multilingual chatbots and domain-specific NLP applications for global businesses. Championed MLOps best practices with Docker, Azure ML, and GitHub Actions, cutting model retraining cycles by 50%.

Designed and deployed end-to-end ML pipelines with Python, Scikit-learn, and TensorFlow, automating feature engineering, model training, and evaluation, which reduced time to production from weeks to days for predictive analytics projects. Developed a recommendation system using PyTorch and LightGBM, integrating it with a RESTful API that served real time predictions to over 500K monthly users, boosting customer engagement by 18%. Implemented NLP pipelines with Hugging Face Transformers, spaCy, and BERT finetuning for sentiment analysis and document classification. Built a data preprocessing and quality monitoring framework with Pandas, Spark, and Airflow, enabling automated anomaly detection and improving data reliability. Collaborated with cross-functional teams to deploy ML models on AWS, ensuring scalability and reducing cloud infrastructure costs through efficient resource allocation. Introduced MLOps practices with MLflow, GitHub Actions, and Kubernetes, standardizing experiment tracking, CI/CD, and model deployment, which improved reproducibility and reduced model downtime

Conducted exploratory data analysis and predictive modeling using Python, Pandas, NumPy, Scikit-learn, enabling data-driven decision-making that improved customer retention forecasting accuracy. Designed and deployed machine learning models with TensorFlow and Scikit-learn, supporting fraud detection and risk assessment use cases that reduced false positives. Built and optimized ETL pipelines with SQL, Spark, and Airflow, automating data ingestion from multiple sources and improving pipeline efficiency by 30%, which ensured reliable data availability for analytics dashboards. Created interactive business intelligence dashboards in Tableau and Power BI, providing real time insights into KPIs such as churn rate, revenue trends, and user segmentation for senior stakeholders. Applied natural language processing with NLTK and spaCy to analyze customer feedback and call center transcripts, uncovering sentiment patterns that drove product feature enhancements. Collaborated with engineering teams to implement A/B testing frameworks with Python and statistical methods, enabling data-backed product rollouts and improving feature adoption rates.

Contributed to web application development using JavaScript, TypeScript, React, and Node.js, gaining professional experience in modern software practices. Assisted in building user interfaces with React while gradually adopting component-based design patterns to improve maintainability. Supported backend logic in Node.js, handling simple API routes and integrating with databases under senior guidance. Practiced TypeScript for stronger typing and reliability, reducing runtime issues in collaborative projects. Applied basic styling with CSS, Sass, and Bootstrap to ensure functional and consistent layouts. Participated in team sprints, standups, and code reviews, developing an understanding of agile workflows and collaborative coding standards. Quickly adapted to production environments, transitioning from academic projects to professional-grade applications.