Anshika Joshi

Agentic AI-Driven Chatbot (Natural Language to SQL)

• Designed and built from scratch an agentic AI chatbot using LangGraph, LangChain, and Gemini LLM to convert natural language queries into SQL, and generate human-readable insights.
• Improved query interpretation accuracy to 90%, boosting decision-making efficiency, and end-user satisfaction.
• Led multi-agent workflow orchestration, tool integration, and scalability design, ensuring system robustness.

Retrieval-Augmented Generation (RAG) for M&A (D-ICE Platform).

• Implemented a RAG pipeline with Milvus and structured dataset alignment to enhance document intelligence for Deloitte’s M&A platform.
• Achieved significant improvements in retrieval accuracy, directly impacting due diligence, and reporting quality.
• Enhanced parsing pipelines with Azure AI Document Intelligence, integrated LLMs for contextual comprehension.
• Applied RAGAS for performance evaluation.

Voice Cloning and Text-to-Speech (TTS) for Indian Languages

• Designed and implemented an end-to-end pipeline for multilingual TTS and voice cloning, with a primary focus on Hindi.
• Collected and processed large-scale audio-text datasets for model training and fine-tuning.
• Developed speaker-adaptive voice cloning models by fine-tuning TTS systems with speaker-specific data, enabling high-fidelity replication of human voices.

AI Chatbot Development (Intent Classification & Entity Recognition)

• Built an AI-powered chatbot, leveraging NLP techniques for intent detection and entity extraction.
• Implemented robust tokenization, preprocessing, and NER models to enhance accuracy in understanding user queries.
• Optimized chatbot pipelines for real-time query classification and contextual entity recognition, improving conversational accuracy and user engagement.

Generative AI – Context-Based Question-Answering System

• Developed a RAG (Retrieval-Augmented Generation) pipeline using LangChain and LLMs for intelligent Q&A from documents.
• Engineered a document ingestion pipeline: parsing PDFs, chunking text, and generating embeddings stored in a vector database.
• Designed context-retrieval mechanisms to fetch relevant embeddings, and integrated them with LLM prompts for precise, context-aware answers.
• Enhanced system performance by fine-tuning retrieval and generation flow, ensuring higher accuracy and relevance in responses.

• Performed in-depth exploratory data analysis (EDA) using Python and R to uncover patterns, detect outliers, and generate actionable insights for business decision-making.
• Implemented clustering algorithms (K-Means, DBSCAN) to segment data, identify hidden structures, and support data-driven strategy development.
• Applied deep learning techniques, including neural networks and autoencoders, for anomaly detection in large-scale datasets, improving accuracy in identifying irregular patterns.
• Contributed to end-to-end product development, designing relational database schemas, and developing scalable backend APIs using the Django Rest Framework (DRF).
• Administered and optimized PostgreSQL databases across development and production environments, ensuring high availability, security, and performance.