● Speech transcript (ASR) from the speech data and extracting the speaker id(Diarization) from the various LLM models.● Developed advanced contextual and knowledge-based chatbot solutions using GenAi LLM, effectively handling both unstructured and structured data. Integrated GenAi with Langchain to manage a comprehensive DB comprising more than 200 KPIs from 70 structured data(Research ,Sales and marketing data), while interacting with a vast 10 GB of unstructured data. Conducted extensive prompt engineering & RAG techniques to optimize the system’s responsiveness and intelligence & deployed GenAi solutions on the client’s VM & Integrated the streaming functionality like Chat GPT interface in solutions, which helps the internal audience/employees/users to understand the Research data much better.● Developed advanced Knowledge based support chatbot solutions using GenAI, RAG, and Langchain tools and managing multiple KPIs from structured data sources and unstructured data (Financial Data). Optimized the system's performance through prompt engineering and retrieval-augmented generation. Deployed GenAI on the client's VM, integrating ChatGPT-like streaming for efficient user interaction with research data, which helps the audience to increase the efficiency 4 times faster.● Finetuned the Llama-2 7B model with LoRA and QLoRA finetuning methodologies on financial data and deployed on Azure cloud ● Developed a pipeline integrating Transformers - based models for Translation, Summarization, Audio transcript, Image captioning, Language detection, and Location extraction/Coordinate extraction/ISO, etc., along with separate multilabel and multiclass NLP models to predict the keyword-matching from news-based data, enabling efficient retrieval and insertion of data from a MongoDB.● Extracted different clauses from the OCR scanned contracts data and fine-tuned the LLM(Q&A) model to predict different Entities.● Interacting with the clients and proposing the architecture

● Collaborated and real-time data collected from Hospitals with the observation of expert Doctors.● Pre-processing and denoising the PPG signal and understanding of physiological changes with the PPG signal (Sensor data)● Various features extracted & predicted various healthcare parameters from PPG Signal using ML models (HR, RR, Glucose level, etc.)● Optimized Deep learning models on various medical image data with 3DUNet, Cancer Net, VGG, and other Transfer learning models● Pre-processed & built the multi-classification machine learning models on physicochemical dynamical simulated data & compared the accuracy and ROC scores which helped to reduce the experimental time in drug discovery by 65%. ● Pre-processed & Built the Artificial Neural Network Regression model to predict the IC50 values from the nano physicochemical dataset which helped to reduce the experimental time in drug discovery by 50%. ● Built the bidirectional LSTM Regression model to predict the protein’s 3D structure morphologies. Which helps to understand the functionality of the organism.● Developed end-to-end ML models for predicting PCOS using both invasive and non-invasive datasets, facilitating quicker patient treatment decisions for healthcare professionals

• Engineered a data pipeline for machine fault detection by scrapping the data frame API and utilizing AWS S3 for efficient storage of raw operational data.• Transformed and structured raw data into useful formats like CSV via AWS Athena using stored procedures, enhancing data readiness for analysis.• Conducted data analysis and visualization using Apache PySpark, enabling comprehensive exploration and insight extraction from machine operation data.• Built and managed a real-time ETL pipeline utilizing Apache Kafka to ingest and process streaming data, ensuring continuous availability of clean data for further modelling and analysis.• Streamlined data workflows, facilitating faster and more efficient data preparation for future machine learning applications

● Developed and implemented advanced data visualization techniques for banking industry data using Tableau and Python to generate business insights. These visualizations facilitated informed decision-making among stakeholders, resulting in a significant 20% increase in actionable insights derived from reports.● Engineered optimized Extract, Transform, Load (ETL) processes using SQL Server Integration Services (SSIS) and SQL queries achieving a 25% reduction in data processing time and improved data accuracy by 15%, enhancing overall data reliability for analysis.● Implemented robust data validation and cleaning procedures, leading to a 30% reduction in data anomalies and errors.● Streamlined and automated reporting processes using SSRS and report-building software which reduced report generation time by 40%, enabling faster access to critical information and enhancing efficiency across departments.