Anish Pati, Ms

• Developed scalable, headless scraping pipelines using Crawl4AI and Selenium WebDrivers to extract consumer reviews across platforms, enabling structured downstream analysis.
• Built a review summarization engine, combining data cleaning, sentiment analysis, and BERT-based classification, to extract pros and cons from consumer and employee reviews.
• Achieved 94% classification accuracy compared to the previous 83% by optimizing a news categorization model through NLP preprocessing, dimensionality reduction, EDA, and rigorous model selection.
• Engineered a financial data extractor using generative AI to parse key financial parameters from PDF-based annual reports, guided by schema-aware prompts and document parsing techniques.
• Created a custom web search and retrieval-augmented generation (RAG) tool using Serper, HyDE, and multi-query synthesis, achieving approximately 80% of Perplexity’s performance at 10% of the cost.
• Designed an agentic AI chatbot framework integrating company data via custom tools (financial, vector databases via RAG, web search), enabling dynamic, multi-turn, task-oriented conversations with high retrieval precision.
• Built a multi-agent financial news analysis system to extract, categorize, and summarize company-relevant insights from real-time data feeds using task-specialized agno.ai agents.

• Served as a solutions architect, optimizing existing predictive methods for various operations for a Philippine oil retailer company.

Led the implementation of predictive models, resulting in a $250,000 increase in annual revenue, and a $10,000 reduction in annual costs.

• Developed prescriptive models using ensemble techniques, combining tree-based and transformer-based models, achieving lower MAE than the individual models.