SHAONI CHAKRAVARTY

Designed & Developed a Multi-Agentic workflow for healthcare with IBM Watsonx products:

• Developed a multi-agent workflow solution for the healthcare domain using IBM Watsonx Orchestrate to enhance user interaction and automate data processing
• Leveraged Large Language Models (LLMs) to detect user intents from utterances, generate responses, and trigger tool calls
• Designed multi-agent flows using Python and explored LangGraph to build scalable architectures
• Integrated responses from multiple agents into downstream workflows, enabling seamless data population across widgets
• This innovative approach streamlined complex processes, improving efficiency and user experience
• Designed and developed a comprehensive terrorist database solution by scraping client websites and integrating open-source datasets
• Utilized Docker for deploying these cloud-native data workflows, ensuring portability and scalability in production environments.
• Maintained robust coding practices and version control through GitHub, fostering team collaboration and high-quality delivery.
• Achieving a 30% reduction in process turnaround times.

AI-Powered Counterterrorism Intelligence System with IBM Watsonx:

• Leveraged IBM Watsonx products to build a conversational chatbot using Watsonx Assistant, enabling users to query detailed information about terrorists
• Integrated a knowledge graph with dynamic UI features to display terrorist profiles, including their image and connections
• Provided actionable insights, such as attacks executed by terrorists and their associated organizations, locations on a world map, weapons used, and casualty statistics
• Enhanced the solution with LlamaIndex for comparative analytics, delivering advanced intelligence and visualization capabilities.
• Provided actionable insights that improved intelligence analysis accuracy by 20%, supporting informed decision-making.

Automated workflow management with IBM Watsonx:

• Automated Trello's project management workflows using IBM Watsonx Orchestrate and LLMs to streamline task management and enhance efficiency
• Developed a solution enabling users to create project dashboards, manage teams, and perform actions such as creating task cards, adding comments, and moving actions between cards seamlessly
• The automation reduced project risk and impact analysis by 30%, improved sprint planning, and provided high-level overviews with minimal human intervention, fostering better collaboration and decision-making
• IBM is revolutionizing AI and automation with Watsonx, a cutting-edge AI platform designed for scalable model training, deployment, and orchestration across diverse enterprise workflows.
• Achieved a 20% reduction in project risk and impact analysis time, improving collaboration and decision-making.

Product enhancement on Watsonx.data

• Enhanced data ingestion and transformation workflows in Watsonx.data by leveraging Apache Iceberg and PySpark, enabling efficient read/write operations and performance tuning of data pipelines.
• Conducted performance tuning and measured latency, achieving 20% faster data pipeline performance.

Tech Stack: Python, IBM Watsonx products, LLMs, LangGraph, Knowledge Graph, LlamaIndex, Web Scraping

Optimizing EV Charging Station Placement Using Advanced Data Modeling:

• Optimizing location of EV charging stations: The project objective is to optimize the location of electric vehicle (EV) charging stations
• The project data includes bounding coordinates, which have been transformed into a JSON file
• It aims to leverage data analysis and modeling techniques (Markov Decision factor, time-series, frequency domain pattern analysis, multivariate forecasting model, probabilistic models) to improve the placement of charging stations and enhance the overall accessibility and efficiency of EV charging infrastructure
• Outcome: Improved charging station coverage by 15% and reduced placement inefficiencies by 20%, contributing to better EV infrastructure planning.

Recommendation model:

• The main objectives were to analyze the average tenure and salaries of employees across different grades and study the compensation (CTC) mix between new hires and lateral employees for each grade
• The project utilized Holt-Winters and SARIMA models to address specific use cases
• By leveraging these models, the project aimed to provide valuable insights and recommendations to optimize hiring strategies and ensure competitive compensation practices within the organization.
• Outcome: Provided actionable insights that improved hiring decisions, aligning compensation practices with market competitiveness and reducing compensation-related attrition by 10%.

• Developed a churn prediction model using predictive analytics for customer retention
• Proposed a classification model to improve risk scoring by reducing false positives
• A leading provider of digital transformation, consulting, and business reengineering services