Pavan N

AI and Machine Learning professional with extensive experience in automating data-intensive processes and delivering AI-driven solutions across diverse sectors, including healthcare, FMCG, oil & gas, and engineering. Adept at leveraging deep learning models, OCR, and document intelligence techniques to drive efficiency and reduce manual effort. Proven track record in successfully implementing solutions that enhance operational workflows and ensure high accuracy in data extraction and analysis.

1. Cavity Detection in Panoramic Radiographs using Deep Learning

• Built a UNET based deep learning model with 78% IoU for detecting cavities in panoramic radiographs

2. Comparison of Digital Packaging Labels (FMCG Sector)

• Developed an AI-driven solution to compare digital packaging labels using entity recognition for text and symbols.
• Integrated rule-based matching and similarity scoring to identify discrepancies.
• Reduced manual quality assurance effort by over 95%, accelerating product label verification.
• Tools: Python, OpenCV, spaCy, Pandas, Scikit-learn.

3. Digitization of Piping and Instrumentation Diagrams (Oil & Gas)

• Automated extraction of Regions of Interest (ROIs), component classification (inlets/outlets), and object detection (valves, pumps, arrows) from engineering diagrams.
• Built custom deep learning models for arrow orientation and flow detection; enabled editable CAD conversion.
• Reduced manual digitization time from months to minutes.
• Tools: PyTorch, OpenCV, YOLO, custom annotation tools, CAD libraries.

4. Engineering Bill of Materials (EBOM) Analysis

• Built a document intelligence system to process multi-page engineering BOMs with GPT-based chain-of-thought reasoning.
• Extracted and restructured tabular data while preserving hierarchical relationships and formatting across pages.
• Applied OCR and rule-based validation for scanned sources; improved conversion accuracy and reduced manual rework by ~30%.
• Tools: Python, GPT APIs, LangChain, Pandas, Tesseract.

5. Spare Parts Recommendation System

• Developed a transformer-based model to predict future spare part purchases from customer history.
• Designed a full ML pipeline for training, validation, and inference; improved top-N recommendation precision by ~20%.
• Tools: PyTorch, Hugging Face, MLflow, FastAPI.

6. Intelligent Document Processing & Retrieval System

• Engineered a platform for semantic document search and structured extraction using OCR, RAG pipelines, and LLMs.
• Enabled batch PDF ingestion, summarization, key-value extraction, and Q&A via a custom Streamlit UI.
• Reduced manual document review time by ~85%.
• Tools: LangChain, OpenAI GPT, FAISS, Streamlit, PyMuPDF, Pandas.

7. Wiring Chart Analyzer

• Built an automated pipeline to extract wire lengths, part numbers, and connection details from multi-page PDF wiring diagrams.
• Computed wire segment lengths based on positional metadata and standardized part references.
• Generated structured Excel outputs for BOM and wire list integration, reducing manual compilation effort significantly.
• Tools: Python, PyMuPDF, Pandas, OpenPyXL.

• Certificate of Recognition for filing a patent at Indian patent Office titled "Method & System of Comparing Digital Labels" (IN202141039416), awarded by L&T Technology Services.
• "Method and System for generating editable schematic illustration of non-editable schematic illustration" (IN202541019135)
• "Method and System for determining pipeline flow in non-editable schematic illustration" (IN202541025242)
• "Method and System for correcting skewness in images" (IN20254103399)

Microsoft Certified, Azure AI Engineer - Microsoft.

• Strong desire to stay updated with the latest advancements in AI/ML technology.
• Actively contributes to team culture fostering diversity, equity, and inclusion.