Transforming over 4 years of R&D into effective product features • Took full responsibility for the end-to-end ML workflow, translating advanced R&D into deployable, high-impact product features • Significantly improved the search engine and recommendation system by implementing LLM and NLP solutions, resulting in a surge of up to 70% in engagement metrics - Achieved up to a 70% increase in user engagement metrics through refined search and recommendation capabilities. - Reduced NLP-based inference latency significantly by implementing advanced optimization techniques such as model quantization, ONNX conversion, and feature store caching - Technical stack: PyTorch, HuggingFace, FAISS, Dask, Elasticsearch, NLTK, Transformers • Designed and deployed an MLOps architecture on Google Cloud Platform (GCP), focusing on scalability and operational efficiency. Key components included: - Configured Vertex AI to facilitate robust machine learning operations, optimizing resource usage and process flows - Integrated Kubeflow V2 pipelines, Kubernetes, and Docker for streamlined deployment and management of ML models - Developed continuous integration and deployment pipelines (CI/CD) using BitBucket pipelines - Technical Stack: Kubeflow V2 pipelines, Kubernetes, Docker, BitBucket pipelines, Flask, GUnicorn

Bootstrap and Scale B2C E-Commerce Project • Outcome: Achieved $2M in revenues over 2018-2020 • Implemented continuous A/B testing • Utilized predictive modeling to boost user acquisition and and Return on Ad Spend (ROAS) - Developed end-to-end machine learning pipelines to forecast advertising revenues versus expenses, employing time-series forecasting and majority modeling - Integrated data from Facebook and Instagram (manager & analytics) APIs, Shopify APIs - Technical stack: Numpy, Pandas, Scikit-Learn, Scipy, PyTorch, PyMC3 • Implemented customer segmentation using Gaussian Mixtures for targeted marketing, enhancing customer engagement • Decreased customer churn by 40% through targeted interventions using supervised learning and Bayesian optimization • Designed and deployed web-based optimization tools to aid the marketing team in strategic decision-making through a man-in-the-loop approach Developed a stock trading bot using advanced time-series forecasting, enhancing intraday trading strategies and decision-making • Crafted trend and volatility predictions by leveraging sophisticated forecasting models - Algorithms used: Long Short-Term Memory (LSTM), Seasonal AutoRegressive Integrated Moving Average (SARIMA), and classification and regression of lag features • Implemented outlier detection to improve the accuracy of technical indicator analysis - Techniques: DBSCAN for clustering, rolling statistics, and supervised classifiers • Advanced quantitative data pre-processing and feature engineering to optimize financial data handling and model performance - Enhanced reliability of Open, High, Low, Close (OHLC) data through outliers capping - Performed large-scale imputation on candlestick and time-series data, ensuring data integrity and usability • Technical stack: PyTorch + TorchServe, Java Spring, FastAPI, Apache Spark, Hadoop, Databricks

Developed the first collaborative software modeling and code generation cloud platform aimed at software developers and architects, significantly enhancing collaborative capabilities in software design • Secured over €1M in funding and received two national innovation awards for breakthrough technology. • Designed and implemented a web-based modeler that streamlined the prototyping process • Engineered robust code transformation and generation pipelines • Technical Stack: Java, Java Spring, GWT, Talend (ETL), AWS (EC2, S3)Led User Acquisition (UA) and Conversion Rate Optimization (CRO) strategies, significantly improving onboarding processes and user engagement: • Applied analytics and A/B testing tools to refine and enhance user acquisition and retention strategies • Technical Stack: Google Analytics, Optimizely, BIME (Zendesk)Mentored a team of over 10 members and led a Scrum agile development team, fostering a collaborative and efficient work environment • Utilized JIRA and Confluence to manage projects and maintain communication across the team.

Designed meta-heuristics to solve large-scale NP-hard optimization problems: • Developed and implemented MPI-based master-slave parallel and distributed evolutionary algorithms • Scaled computational models to GRID’5000, enhancing processing efficiency by over 20%, which significantly accelerated research outcomes. (Stack: C++, CMake, GCC, Debian, Bash)Clean code evangelist; mentored 6 software engineers in best practices and code efficiency.Engaged in software engineering research focusing on meta-modeling and code generation. Published findings in international conferences and journals.