Designed and implemented a Human-in-the-Loop (HITL) financial query agent using LlamaIndex workflows and the FastAPI WebSocket architecture. The system intelligently parses natural language queries to extract financial parameters, dynamically requests missing information through interactive clarification steps, and retrieves precise financial data from structured databases. This agent enhances user experience by guiding users through complex multi-parameter queries while maintaining real-time communication and ensuring data accuracy through human validation checkpoints.

Implemented an SQL agent to automate database queries and streamline data retrieval processes. This agent is designed to handle dynamic query generation, allowing for efficient interaction with complex databases. By automating these interactions, the SQL agent provides timely insights and facilitates better data management, enhancing overall operational efficiency.

Implemented advanced document summerization techniques to extract key insights from large texts. This system leverages AI algorithms to condense lengthy documents into concise summaries, enhancing the accessibility and usability of information. By automating the summer‑ ization process, users can quickly grasp the main points of complex documents without the need for extensive reading.

Developed a question‑answering system capable of extracting accurate responses from CSV datasets. This system allows users to query large CSV files and receive precise answers based on the data contained within. By enabling efficient information retrieval, the solution helps users to quickly find and utilize the data they need, improving decision‑making processes.

Feed Recommendation system Video similarity using feature embedding extraction. Session based recommendation system. Text content based recommendation system using feature embedding extraction. The pipeline for model training and model providing was developed using APIs.

Product Recommendation System Collaborative based recommendation. Content-Based Recommendation. Knowledge-based recommendation system Incremental learning system designed. Incremental training based on frequency of new user or item added.

Proficient in designing end-to-end conversational AI solutions using Rasa frame- work. Expert in NLP for intent recognition, entity extraction, and sentiment analysis. Skilled in crafting contextual dialogue management systems with Rasa’s Core frame- work. Experienced in training and optimizing NLP models for improved accuracy. Successfully deployed and integrated Rasa chat-bots on various platforms. Designed custom actions and policies for executing complex tasks. Continuous improvement of intent and entity recognition. Implemented analytic and monitoring for performance tracking. Thorough testing and quality assurance for reliable functionality.

Automated Job Description Generation using GPT-2 Model. Data gathering for Collecting and pre-processing data-sets for the job. Customise the language model for coherent job postings when fine-tuning GPT-2. The user interface for Develop a user-friendly user interface for simple generating. Gather feedback and make incremental improvements as part of continuous train- ing

Named Entity Recognition using spacy and CRF. Entity Extraction from Candidate Resume. Training Custom data set for getting candidate skills. Optical character recognition. easyOCR, Pytesseract, TrOCR. Whats app automation using selenium.

These projects offer a rich array of experiences. From creating a COVID‑19 chatbot with Rasa for vital information dissemination to conducting sentiment analysis on airline data through a Streamlit app, they hone conversational AI and data analysis skills. Building a Flask‑SQLite web app for URL shortening provides hands‑on application development expertise. Crafting the K‑Nearest Neighbor algorithm from scratch enhances machine learning proficiency. Lastly, participating in a FIFA 19 dataset hackathon for EDA and applying clustering techniques like K‑Means and PCA deepens insights into unsupervised learning, making for a well‑rounded skill set.

have extensive experience in machine learning and data analysis, having implemented 10 different classification techniques for leaf classifica‑ tion, conducted heart attack analysis and prediction using SVM, Random Forest, and logistic regression, and built a simple linear model from scratch with gradient descent, and successfully predicted house prices using advanced regression techniques.