Prathamesh Palkar

Project Experience:

Generative AI initiatives (Jan 2023 – present)

(Microservices, Azure OpenAI, Azure AI Document Intelligence, .NET, Python)

• Document Smart Attributes: Expanded a Generative AI–based tagging system to recognize and categorize content from PDF, DOCX, and CSV files, resulting in efficient mapping to internal workloads.

• Conversational AI (RAG Architecture): Led the development of a Retrieval-Augmented Generation solution, enabling Deloitte Tax professionals to query and interact with tax return PDFs in real time.

• Generic Data Extractor: Designed and implemented a .NET/Python generative AI tool to parse PDF documents (invoices, tax forms, etc.) And extract key data points to convert unstructured data into predefined output templates.

XTRACT K1 ML (Jun 2020 – Present)

(C#, Python, Machine Learning, NLP)

• Built a Windows service to parse tax return PDFs and extract key information for filing.

• Developed binary and multi-class text classification models (scikit-learn) to segment unstructured text by form type.

• Implemented a Named Entity Recognition module (Flair + domain heuristics) to accurately identify critical data points.

• Created a Flask-based API to operationalize NLP models, interfacing with the Windows service and storing results in MS SQL.

PPI Coding & Reporting (Jun 2018 – Present)

(Alteryx Designer, Python)

• Developed Alteryx workflows to assign PPI codes to current-year inventory items based on prior-year data.

Cleaned, filtered, and summarized inventory line items, ensuring data integrity.

• Built a multi-class text classification model (Python/scikit-learn) to automate code assignment based on features such as description length and n-grams.

• Combined model predictions with heuristic rules in Alteryx to achieve high accuracy PPI code mapping.