SHALINI DHOTE

Spearheaded development of a Generative AI-powered chatbot using Vertex AI, integrating RAG pipelines for context-aware query resolution on pull requests. Designed and implemented CI/CD pipelines with GitHub Actions, ensuring automated LLM fine-tuning, deployment, and version control for iterative model improvement. Engineered an Agentic AI-based Phishing Attack Detection System using LangGraph and Autogen, enabling multi-agent collaboration for detection, analysis, and response. Orchestrated MLOps workflows across GCP and AWS, enabling scalable deployment, monitoring, and management of advanced LLM-driven models in production.

Worked with Medidata AI Lab at Dassault Systmes SE, a leading French software corporation known for its advancements in 3D product design, simulation, manufacturing, and other 3D-related technologies across marine, aviation, healthcare, and life sciences domains. Developed an end-to-end machine learning model to predict clinical trial enrollment rates, enhancing efficiency and accuracy in trial planning and execution. Collaborated daily with the New York team on research and innovation, leveraging AI to drive advancements in clinical trials and Intelligent Trials methodology. Implemented Generative AI and LLM to create synthetic data for clinical research, ensuring data privacy compliance and expanding datasets for training without compromising patient confidentiality. Optimized machine learning pipelines with Docker, SQL, Python, R and AWS Batch for streamlined deployment and efficient data processing, significantly enhancing model performance and reducing development time. Worked on the disease diagnostics project, focused on detecting and classifying diseases from radiographic images using AI, Computer Vision and deep learning techniques to improve diagnostic accuracy and efficiency.

Anthem Inc, USA: Built a framework for checking the effectiveness of the US healthcare program using Causal Inference Method. Built a statistical model and found correlation between independent variables for explanatory data analysis problem. Built Random Forest Regressor model to calculate propensity score for both the Treatment Group and Control Group. Used K-Nearest Neighbor algorithm for propensity score matching. Calculated Inverse Propensity Weighting for both groups to assess balancing using Standardized Mean Difference. Created second stage regression model to calculate Difference-in-Difference and ATE. T-Mobile Inc, USA: Built Machine Learning Model for bandwidth classification using Decision Tree, Naive Bayes classifier. Wrote SQL query to fetch signal data from database. Anthem Inc, USA: Built custom NER-NLP model for Digital Nurse Assistant which helps the nurses by highlighting the most important keywords of DCN (patient prescription) that were taken from the policies provided by the US Healthcare department.

Edvak Inc: Spearheaded a Generative AI Conversational Voice Assistant Team of 6 teammates. Built a chatbot and voice bot that has functionalities such as booking an appointment, no-show ML prediction model, taking vitals of patients. Implemented blender bot, medical BERT, LLM, NLU preprocessing, intent classification, NER, topic modeling, semantic classification, and aspect-based sentiment analysis in the bot. Broadridge Inc, USA: Worked with New York based audit team to automate tasks using python, AI/ML. Executed end-to-end development of production level python algorithm to build anomaly detection model to check how many employees are claiming false expense reimbursement. Built dashboards using Plotly-dash library and Power BI and worked with Knime tool for model building. Implemented Time Series Analysis using ARIMA model to check the trends of vulnerabilities opening in the next three months. Created Sentiment Analysis model to classify incident summary as high, low, and critical priority using NLP, Random Forest, Naive Bayes Classifier. Groofl Intellect: Developed end-to-end Automated AI trading robot for algorithmic trading. Built a model for stock price prediction using Yahoo finance API for finance data. Created an ML model for sentiment analysis on top ten news headlines of stocks. The AI robot suggests when to buy and sell stocks. Deployed model on AWS using EC2 and SageMaker.