Ajay Swarnkar Email & Phone Number

Bhopal, Mp, In

Mumbai, Maharashtra, India

- Developed CHATBOT using RAG, LANGCHAIN: 1. Used the Retrieval-Augmented Generation (RAG) technique to retrieve the most relevant documents as per the user’s query.2. Utilized the textembedding-gecko model to create embeddings and matched using Google’s Matching Engine.3. Worked on Langchain, ConversationalRetrievalChain, FastAPI, GCP Cloud Run, Embeddings’ LLM Models.- NATURAL LANGUAGE TO SQL QUERY using Gen AI LLMs: 1. Used the CodeGeneration model to generate the SQL query from natural language to get the desired results from the BQ table.2. Worked on TextGeneration model, gave natural language with fetched results as context to get answers in a human-readable form. 3. Did thorough Prompt Engineering on both models to carve the desired results from the LLM.- CONTRACT REDLINING Solution for a Catholic Hospital Network using LLMs, RAG:1. Redlining Contracts between two parties based on the clauses provided by the client.2. Used Elastic Search to store Embeddings of the clauses extracted from the given contract as input.3. Working on Fine-tuning Google's Palm2 textbison model to do the redlining on the contracts as per the clauses (to add some missing clauses, to remove some extra clauses, or to modify the existing clauses as per the instructions by the client). Did Supervised Fine-tuning (SFT) of the textbison model on GCP by creating a manual dataset in JSONL format.- PROCURE TO PAY using OPENAI Models:1. Built a solution for automating the procurement request process using GEN AI.2. Used OPEN AI Function Calling to act as an API Selector and run the desired the functions.3. Utilized AzureChatAPI, GPT Models to automate the process wherever required.

Mumbai, Maharashtra, India

1. TIME-SERIES FORECASTING - Predicted the total depletions of the client’s products for the future to be allocated across the states of the US. Built Time-series forecasting and Regression models on GCP Vertex AI, and did the Batch Predictions.Built end-to-end MLOps Pipeline on the complete workflow.2. CHURN PREDICTION - Developed a Neural Network (NN) based solution to find whether a particular customer will be retained or churned. Used the RFM Model technique to segment the customers.Visualized the data on a Web UI developed using Streamlit.Deployed the model with ~81% accuracy and identified the customers for businesses to do marketing campaigns.3. RECOMMENDATION SYSTEM - Build a solution for finding the following best product for a customer using the Two-Tower Model (Tensorflow Recommenders).Worked on User/ Item Embeddings, built a model to predict the top n products for a particular customer to increase revenue, develop product awareness and customer relations.4. ASPECT-BASED SENTIMENT ANALYSIS (ABSA) - Developed a solution for Classification of accounts (client’s product selling stores) based on customer reviews. Used customer reviews to identify the poor-performing accounts, and implement market strategies (by business).Worked on text augmentation, XGBoost, Word Embeddings, Lemmatization, and TFIDF.Increased sales of their products by 7.39%.

Mumbai, Maharashtra, India

Completed the training program successfully and understood all the concepts of Machine Learning, Deep Learning, Deployment tool, various services of Google Cloud Platform, etcShadowed in a Churn prediction project, researched about different models that can be used for Churn Modelling, Customer retention.

Bachelor Of Technology - Btech, Information Technology, 8.71/10

Hsc Boards Exam, Maths Science, 90%