Prathamesh Palkar

Generative AI initiatives (January 2023 – present)

(Microservices, MongoDB, Azure OpenAI, Azure AI Document Intelligence, .NET, Python, Langchain, LangGraph, OPIK)

• Agentic AI Tax Return Preparation: Implemented a multi-agent architecture via LangGraph to automate individual tax return workflows, significantly reducing manual intervention.
• Document Smart Attributes: Led the development of a Generative AI–driven tagging system (PDF, DOCX, CSV), improving classification speed by 30% and streamlining workload mapping.
• Conversational AI (RAG Architecture): Built a Retrieval-Augmented Generation solution, enabling real-time user interactions with tax return PDFs; introduced LLM observability (OPIK) to monitor answer relevance, contextual precision, and hallucination rates.
• Generic Data Extractor: Architected a .NET/Python-based tool for parsing diverse PDF documents, including invoices, tax forms, etc. into standardized templates, slashing manual data extraction by 80%.

LLM Metrics Reporting (January 2024 – present)

(MS SQL, Power BI)

• Developed a Power BI dashboard to track tokens per minute utilization (TPM) and PTU usage for LLM applications, providing actionable insights for cost optimization.
• Identified opportunities to reduce PTU provisioning and TPM usage for substantial cost savings.

ESG Benchmarking (October 2022 – October 2023)

(Azure Databricks, Python, Topic Modeling, NLP, Power BI)

• Built a large-scale data pipeline in Azure Databricks to ingest and transform SEC Edgar 10-K filings.
• Leveraged topic modeling and NLP to generate ESG insights, integrating processed data (Delta Tables) with Power BI for comprehensive benchmarking.

XTRACT K1 ML (June 2020 – Present)

(C#, Python, Flask Framework, Machine Learning, NLP, Microservices)

• Created microservices to parse tax returns, achieving 95% accuracy in classifying form types via binary and multi-class ML models.
• Deployed Named Entity Recognition (Flair + domain heuristics) to pinpoint critical data fields.
• Developed a Python API to operationalize the models for seamless integration with .NET services.

SalesEdge (C# ASP.NET, MS SQL Server, WCF)

• Built a web application for a major U.S. tobacco manufacturer to manage retail payments, promotions, and employee assignments.
• Created a bulk payment approval module for executives, cutting manual effort, and streamlining approval workflows.
• Optimized code and SQL queries to reduce screen render time from 4 minutes to 30 seconds, significantly boosting performance.