Senior AI/ML Engineer with strong hands-on experience building and deploying production-grade machine learning and Generative AI solutions across cloud platforms. Proven expertise in developing end-to-end ML systems from…
Senior AI/ML Engineer with strong hands-on experience building and deploying production-grade machine learning and Generative AI solutions across cloud platforms. Proven expertise in developing end-to-end ML systems from messy, unstructured data ingestion to model training, evaluation, and scalable deployment using Python, Azure, and AWS. Deep experience in classification systems, traditional ML, NLP, RAG pipelines, and LLM-based workflows that replace manual, document-heavy processes with intelligent automation. Known for strong ownership, 0-to-1 execution, and close collaboration with non-technical domain experts to deliver measurable business impact.
Worked closely with business teams to gather requirements and design scalable AI-powered solutions, leading to faster prototype delivery, improved business alignment, and higher project success rates.
Conduct feasibility studies and build Proofs of Concept with AutoGen and CNN/RNN models in Python, confirming technical viability and reducing project risk.
Designed and implemented end-to-end AI solutions using Generative AI, Deep Learning, and Python, leading to faster data analysis and improved predictive accuracy for business decisions.
Tested and validated solutions on real-world sample files, improving accuracy and reliability and helping reduce errors by 20%.
Debugged and resolved errors across data pipelines, LLM workflows, and deployed environments, improving system reliability and reducing downtime.
Deployed applications into client-specific environments, ensuring compliance with enterprise standards and seamless integration.
Worked closely with business teams to gather requirements and design scalable AI-powered solutions, Conducted feasibility studies and built Proof of Concepts with AutoGen and CNN/RNN models in Python, Designed and implemented end-to-end AI solutions using Generative AI, Deep Learning, and Python, Tested and validated solutions on real-world sample files, Debugged and resolved errors across data pipelines, LLM workflows, and deployed environments, Deployed applications into client-specific environments
Developed an automated document classification and information extraction system using machine learning, reducing manual review time by 30% and increasing data accuracy.
Designed and implemented machine learning models that improved classification F1 score by 12% and resolved complex data management issues, leading to faster and more accurate business decisions.
Utilized advanced data cleaning techniques, including tokenization and normalization, to prepare datasets for analysis.
Designed AI-driven solutions using GPT-4.0, Llama3, and Gemini Pro, achieving enhanced accuracy in complex tasks.
Integrated AI applications with multiple LLMs and various data sources using LangChain and Llama-Index, improving real-time data access and reducing integration time by 40%.
Collected and integrated diverse data sources using Pinecone and ChromaDB, increasing training dataset size by 40% and improving model performance.
Implemented prompt engineering and evaluation frameworks, including RAGAS, to ensure reliable outputs and measure solution performance.
Developed and deployed serverless architectures in AWS and Azure for scalable and efficient AI solutions.
Developed an automated document classification and information extraction system using machine learning, Designed and implemented machine learning models to improve classification F1 score and resolve data management issues, Utilized advanced data cleaning techniques, Designed AI-driven solutions using various technologies, Integrated AI applications with LLMs and data sources, Collected and integrated diverse data sources, Implemented prompt engineering and evaluation frameworks, Developed and deployed serverless architectures in AWS and Azure
Developed an AI-driven solution using GPT-4o and Llama-Index to automate entity extraction from invoices across varying templates and geographies. • Automated data categorization and structured extracted data into relational tables, ensuring high accuracy and scalability. • Built a logging mechanism to capture ambiguities and flag exceptions for manual review, enhancing data reliability. • Conducted extensive testing and implemented backup/recovery strategies to maintain system reliability under high invoice volumes
Developed a GenAI-driven predictive and preventive maintenance system for MRI devices to address potential issues and reduce downtime preemptively. • Built a machine learning model to identify anomalies and deviations in MRI system data, enabling early detection of failures. • Developed a RAG-based chatbot to help maintenance teams diagnose anomalies and fix component issues independently, reducing downtime. • Integrated backend functionalities with Angular UI, collaborating closely with frontend developers for a seamless user experience. • Debugged and resolved system errors, ensuring high reliability and stability.
• Developed an AI-driven system with 95% accuracy for the Integration of XML in legal documents (XML) for crime-case judgment content. • Built a modular Python architecture with utility functions for XML parsing, LLM integration, and instruction classification, ensuring extensibility and maintainability. • Built a Python pipeline to extract targeted content sections from final LLM outputs, and applied regexbased rules to parse XML files. • Implemented prompt-driven processing pipeline using customizable templates for instruction parsing and LLM-based content processing. • Created a Streamlit application to orchestrate the complete workflow, enabling end-users to trigger all steps (Instructions type – process – generating outputs) via a simple UI processing button. • Implemented robust debugging, error-handling, and iterative testing to ensure accuracy and production readiness. • Created an automated file management system with an organized directory structure for input/output handling and batch processing capabilities.
Developed an AI-driven system with 95% accuracy to verify star pagination in legal documents (PDF, XML, Excel) for crime-case judgements content. Converted PDF pages into images and applied multi-step LLM processing: a. Extracted content types per page using annotated reference images in LLM run1. b. Corrected continuity of outputs using Python logic. c. Used the corrected output along with original images to extract structured content for all identified content types in LLM run2. d. Re-processed only erroneous pages with additional LLM run3 to enhance accuracy. Built a Python pipeline to extract targeted content sections from final LLM outputs, and applied regexbased rules to parse XML files. Utilized Excel metadata (page numbers, citation IDs) to cross-verify consistency between PDF and XML content. Created a Streamlit application to orchestrate the complete workflow, enabling end-users to trigger all steps (PDF → XML → Excel → Verification) via a simple UI processing button. Implemented robust debugging, error-handling, and iterative testing to ensure accuracy and production readiness.
• Developed a system to classify and segment legal case documents, streamlining access to categoryspecific information for efficient case organization. • Leveraged Azure Form Recognizer to extract structured data, including text, key-value pairs, and tables, supporting accurate document categorization. • Implemented a page-level classification model using First Page Prediction (FPP) for precise categorization of pages into classes like invoices and medical reports. • Enhanced classification accuracy by employing XGBoost and TFIDF vectorization in machine learning models. • Integrated Gen AI and rule-based pipelines for advanced information extraction, tailoring retrieval based on document class, enhancing data utility and relevance for case management.
Developed dual-stage ML system: a. Stage 1: Document classification using Logistic Regression + custom feature extractors (TF-IDF, Doc2Vec, LDA topic scores). b. Stage 2: Paragraph-level importance scoring using hybrid rules + BERT classifier to rank content as High/Med/Low priority. • Integrated text extraction with layout preservation using PyMuPDF and custom XML tag mapping to retain document hierarchy. • Tracked all experiments using MLflow, auto-logged hyperparameters, metrics, and artifacts; enabled versioned deployment on AWS SageMaker. • Implemented chunking strategies (overlapping windows, sentence-level sliding) and prompt tuning for better response coherence and retrieval accuracy. • Created a role-based analytics dashboard using Streamlit for visualizing classification summaries and a user feedback loop. • Business Impact: Cut document review time by 40%, improved team compliance by 2x, and enabled real-time Q&A across 50,000+ technical docs
The AI-first Recruitment Operating System that unifies and automates the entire hiring lifecycle for modern enterprises worldwide.
Dr Job FZ LLC. 2026 © All Rights Reserved
