Design and develop cutting-edge enterprise artificial intelligence solutions within the Azure ecosystem, focusing on digital transformation and process optimization through advanced cognitive technologies. Key Responsibilities: PyTorch: Development of end-to-end training pipelines for NLP and computer vision models using PyTorch Lightning. Implementation of fine-tuning strategies for pre-trained models (BERT, RoBERTa, ResNet, EfficientNet) with custom learning rate schedulers and early stopping callbacks. Model optimization using quantization and pruning techniques. Production deployment with TorchServe and ONNX export for cross-platform inference. Langchain, Langgraph & Langsmith, Tensorflow. Generative AI: Architecture and implementation of applications powered by generative models (Azure OpenAI Service, LLMs) for content automation, intelligent virtual assistants, and code generation, achieving significant improvements in enterprise productivity, some projects are about AI OCR, extract information from documents. AI Agents: Building autonomous multi-agent systems using Azure Bot Service and Cognitive Services, capable of complex decision-making and executing automated workflows with minimal human intervention. Natural Language Processing: Development of robust NLP pipelines for sentiment analysis, entity extraction, automatic translation, and contextual understanding using Azure Language Services and custom models. Computer Vision: Implementation of computer vision solutions for object detection, facial recognition, advanced OCR, and real-time image/video analysis using Azure Computer Vision and Custom Vision. Intelligent Information Extraction: Design of intelligent ETL systems combining Azure Document Intelligence, Cognitive Search, and ML techniques to process unstructured documents, extracting critical business insights with high accuracy. Core Technologies: Azure AI Services | Azure Machine Learning | Python | LangChain | Vector Databases | RAG Architecture | MLOps | Docker | Kubernetes

ProjectBoxcar AI Agents: various technologies: Langchain and Microsoft Autogen. Tensorflow and PyTorch. Microsoft Azure and Semantic Kernel. Strands Agents. AWS Bedrock AgentCore. Artificial intelligence/Machine Learning projects: Polymer Formulation with Artificial Intelligence. Localizing people using WiFi fingerprint: Supervised learning models are used for indoor localization using Wi-Fi fingerprints, covering 110m² of space. KNN, Random Forest, Decision Tree and SVM are used for classification and regression, with feature selection using variance thresholding or PCA. Genetics. Cybersecurity: Detecting anomalies in user activity using convolutional neural networks (CNN), recurrent neural networks (RNN) and Nvidia Transformers. Natural Language Processing. Generative AI: HuggingFace, OpenAI, Llama 3, Claude, Gemini, Fine tuning, RAG. Optical Character Recognition (OCR) to convert diverse documents into searchable and editable formats, improving data accessibility. Elasticsearch: Vector Search, Deploying NLP models, Image Search, Retrieval Augmented Generation (RAG) with Elasticsearch, Web Crawler, Machine Learning with Elasticsearch. Azure Machine Learning/Data Science. AWS Machine Learning/Data Science. NVIDIA: Morpheus, Triton, pipelines. Alteryx. Parallel computing: CUDA Nvidia Machine Learning, DASK. Kubernetes.

Generative AI System for Psychology Company. The conversation between the therapist and the patient is transformed into text, then goes through the text cleaning and normalization pipeline. It is subsequently ingested by the model through RAG techniques. Likewise, specialized psychology texts are uploaded to the LLM and finally with the help of Prompt Engineering, the system is capable of recognizing the problems that the patient may have, as well as hidden information for psychologists to help them design a therapy. The system can also, with the help of specialized texts, design an individualized therapy for the problems detected in the conversations for each patient. Python, Langchain, OpenAI and Streamlit among others were used. With the help of a database of cases, the system's accuracy tests were carried out, resulting in more than 95% acceptable.

Development of an Intelligent Document Processing (IDP) system for automated matching of Bill of Lading (BOL) and Proof of Delivery (POD) documents using Azure AI services. Key Responsibilities: AI-Powered Document Matching: Implemented multi-stage matching pipeline using Azure OpenAI (GPT models) for intelligent similarity analysis between BOL and POD documents with configurable confidence thresholds. Multi-Stage Matching Algorithm: Designed 5-stage strategy including exact BOL matching, AI-based similarity analysis, order number matching, same document type matching, and email ID postprocessing. Azure Functions & Serverless: Built scalable HTTP-triggered Azure Functions for batch document processing with robust error handling and blob storage integration. NLP & Semantic Analysis: Developed algorithms to compare document fields (fuel types, amounts, dates) using semantic understanding for terminology variations. Enterprise Integration: Built integration with Entinuum ERP for order validation with date similarity calculations and vendor matching. Data Engineering: Created ETL pipelines for type conversion, field normalization, and exception categorization.

The following tools and technologies, including Artificial Intelligence, Machine Learning, Generative AI, Transformers, LLM models, ChatGPT/GPT/OpenAI, LangChain, LlamaIndex, Guardrails (Pydantic and Microsoft Guidance), Azure OpenAI, Microsoft Autogen, Vector Database, Retrieval Augmented Generation (RAG), Data Scientist, Natural Language Processing (NLP), Python, Azure, and Spacy, were instrumental in executing a range of innovative projects: 1. Tree of Thoughts Paper (ToT) implementation with Langchain and Python. 2. Microsoft AutoGen implementation (MatchChat, RetrieveChat, Compression, Dalle and GTP4, RAG, Assistant function call, Panning, Completion). 3. LLamaIndex implementation different use cases (Question and Answering, RAG, Summarization, Custom, etc). 4. Chat Analytics: Developed an interactive SQL, CSV, XLSX database interface capable of responding to various user queries. 5. Custom Chat with Langchain and OpenAI: Designed a versatile chat application capable of ingesting and processing large volumes of documents stored in memory. 6. Summarization App Using LangChain and OpenAI: Created a chat-based summarization tool that effectively condenses extensive documents. 7. GPT-4 Automatic Code Reviewer: Engineered an interactive code review system that analyzes and provides feedback on code submissions. 8. Sentiment Analysis LLM OpenAI GPT: Developed a Python-based application to assess and categorize comments on blogs as negative, neutral, or positive. 9. Question and Answering with Hugging Face: Created a versatile Question and Answering system capable of addressing inquiries on various topics.

The following tools and technologies, including Machine Learning, Artificial Intelligence, Natural Language Processing (NLP), Computer Vision, Data Scientist, Python, Microsoft Azure, Spark, Transformers and LLM models, Data Engineering, Spacy, and NLTK, have played a pivotal role in driving the successful completion of the following projects: 1. Object Detection (Computer Vision): Developed a sophisticated system for the detection of objects within images and videos, specifically focusing on identifying objects on ceilings and walls in a factory environment. This project utilized images captured by drones and harnessed the power of Python, OpenCV, Yolo, and Microsoft Azure to achieve precise object recognition. 2. Oil Spill Detection (Computer Vision): Designed an innovative solution for the early detection of oil spills using computer vision techniques. The project leveraged Python, OpenCV, Yolo, and Microsoft Azure to swiftly identify and respond to environmental hazards. 3. Custom Meeting Chat: Created a transformative chat application that converts meeting voice recordings into text transcripts. This text data is then processed through Langchain and OpenAI, enabling the chat system to respond to inquiries and provide insights regarding the meeting content, fostering enhanced communication and information retrieval.

Machine Learning • Artificial Intelligence • Natural Language Processing (NLP) • Computer Vision • Data Scientist • Python • Spark • Data Engineering - Designing and developing NLP models and applications. - Choosing appropriate and effective text representation techniques and algorithms for a variety of task types. - Training, evaluating, scaling, deploying and maintaining models - Design and understand the different processes in information processing. - Program the different solutions for the transformation and analysis of information. - Document the different processes that are carried out for the treatment of information.

The utilization of cutting-edge tools and technologies, encompassing NLP, Machine Learning, Deep Learning, AWS, Named Entity Recognition (NER), Tokenization, Stemming and Lemmatization, Bag of Words, Sentiment Analysis, Sentence Segmentation, Text Summarization, Text Classification, Keywords Extraction, Hugging Face, and Semantic Extraction with TensorFlow, Keras, and Scikit-Learn, coupled with data cleaning and scraping, monitoring, and maintenance of NLP systems, has driven the successful execution of the following projects: 1. Matching Resume with the Job Description: Engineered a system that quantifies the percentage of alignment between a candidate's resume and the job description of the position they're applying for. This project, employing NLP, Python, Spacy, NLTK, and AWS, among others, provides valuable insights to recruiters, streamlining the hiring process. 2. Chat with the Resume: Developed an interactive chat application that loads a candidate's resume, allowing recruiters to ask questions related to the resume's content. This enables recruiters to gain deeper insights into a candidate's qualifications, experience, and education. The solution leverages NLP, Python, Spacy, NLTK, and AWS, among other technologies. 3. Matching Job Descriptions to Candidate Resumes: Utilized machine learning to identify the top three job descriptions that align most closely with a candidate's resume. This data-driven approach enhances the efficiency of matching candidates to suitable job opportunities. The project incorporates NLP, Python, Spacy, NLTK, and AWS, among other tools. 4. Identifying Top Candidates for a Job Description: Leveraged NLP techniques to select the best 100 candidates from a database based on their compatibility with a given job description. This data-driven approach streamlines the candidate selection process, utilizing Python, Spacy, NLTK, and AWS.

1. PPG Heart Rate FacePPG: Implemented a machine learning-based system for heart rate measurement using facial photoplethysmography (PPG) data. Leveraged Python and machine learning techniques to extract valuable insights from facial signals. 2. Credit Risk Assessor: Developed a credit risk assessment model using machine learning and Python, aiding in the evaluation of potential borrowers' creditworthiness. 3. Natural Language Processing for Stocks News Analysis: Employed NLP to analyze news articles related to stock markets, facilitating the assessment of credit risk. 4. BERT-Question and Answer System: Created a question-answering system using BERT and NLP, enhancing information retrieval and comprehension. 5. Face Recognition: Designed a comprehensive facial recognition system capable of identifying individuals in images, videos, and webcam feeds, enhancing security and identity verification. Implementation of facial recognition system using PyTorch for training convolutional networks (ResNet, VGGFace) and facial embedding extraction with ArcFace loss function. Built custom data loaders and augmentation pipelines using TorchVision transforms. 6. Fine-tuning BERT with Hugging Face: Enhanced the performance of BERT-based models using fine-tuning techniques, optimizing their performance for specific tasks. 7. Predicting Stock Prices: Leveraged Python and machine learning to develop models for predicting stock prices, aiding investors in decision-making. 8. Skin Cancer Detection (Melanoma): Designed a medical image analysis system utilizing computer vision, TensorFlow 2, and deep learning to identify and classify skin lesions, particularly focusing on melanoma detection. Medical image classification using PyTorch with transfer learning from pre-trained EfficientNet and DenseNet architectures. Implemented custom loss functions for handling class imbalance and gradient-weighted class activation mapping (Grad-CAM) for model interpretability. 9. Chest X-Ray Medical Diagnosis: Employed computer vision techniques with Keras and TensorFlow to facilitate medical image analysis for chest X-ray diagnosis. 10. Landmark Detection & Tracking (SLAM): Implemented Simultaneous Localization and Mapping (SLAM) using deep learning and image processing, contributing to robotics and autonomous navigation. 11. Brain Tumor Auto-Segmentation for MRI: Utilized computer vision techniques and deep learning with TensorFlow to automate brain tumor segmentation in MRI images, aiding medical diagnosis. 3D medical image segmentation using PyTorch with U-Net and attention-based architectures. Implemented custom volumetric data loaders and augmentation strategies for MRI scans.

As a Python Developer/Data Scientist, I embarked on a journey of data exploration and predictive analytics, combining my passion for Python with a robust skill set. My role demanded a diverse set of proficiencies, from SQL and Pandas to SKLearn, all backed by a solid foundation in the fundamentals of Statistics and Machine Learning. One of my standout achievements was the 'Image Captioning' project, a fascinating foray into Computer Vision. Leveraging Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). In the realm of Natural Language Processing, I took on the 'Machine Translation' project. By building a deep neural network with NLTK and Python, I created an end-to-end machine translation pipeline that could seamlessly convert English text into French. With 'Speech Recognition,' I developed an end-to-end automatic speech recognition (ASR) pipeline. It could transcribe raw audio into spoken language, aligning perfectly with the requirements of the job description. This project spotlighted my contributions to the field of Natural Language Processing. 'Predicting Breast Cancer' was a project that tested my mettle in Machine Learning in R. I engaged in model evaluation, including Ensemble Methods, and achieved an impressive 98% accuracy rate. Throughout my tenure, I also honed my data visualization skills using Matplotlib and Seaborn, created compelling narratives through strong presentation and storytelling, and gained a familiarity with Dimensional Modeling techniques. In summary, my experience as a Python Developer has equipped me with a rich toolbox of skills and a versatile mindset, perfect for tackling the challenges posed by the dynamic world of data science and predictive analytics. I am excited to bring this knowledge and expertise to future endeavors and continue pushing the boundaries of what is possible in the field.

Analysis, design and programming of applications that interact with the SQL database. Translate advanced commercial analytical problems into technical approaches to provide viable recommendations. Create examples, prototypes and demonstrations to help management better understand the data. Design ETLs with SQL. Use of mining, the improvement of the data collection processes and the cleaning of the data used for the analysis.

Use C Language. Analysis and coding of applications in the area of scientific visualization (OpenGL), processing CT images. Parallel programming Clusters (MPI).

Design and implement Java applications. Perform analysis, programming, testing and debugging software. Identification of application problems.

Analysis and development of business systems, such as accounting, accounts payable, accounts receivable, bank management, financial reports, human resources and production management systems, always using SQL database technology and UNIX / LINUX operating system.