Preeja Babu

• Designed and developed Skybot, a Retrieval-Augmented Generation (RAG) system, to analyze aviation documents through a chat interface, reducing analysis time by 50% using Langchain, Azure ai search, Azure Openai GPT4-0 and Llama 70B model. • Built a custom prompt evaluation pipeline for data extraction using Multi-Modal Model(GPT40), Azure Databricks, MLflow and Azure openai. Improving the accuracy by 20 % and reducing the development time by 40 %. for the text classification and data extraction. • Created a document classifier model for categorizing aviation documents into IATA classes using sentence segmentation and BERT, achieving 75% accuracy with a majority vote rule. • Using Promptify and the KOR libraries, an effective LLM-based Named Entity Relation Extraction system was created to extract key-value pairs from unstructured aviation text data. The objectives of this system are to decrease manual labour, enhance data accuracy, and automate the extraction process. Concepts of vertical scaling were employed to get around token limit issues. • Develop data models to transform raw data into meaningful insights and coming up with innovative solutions to be used by the Marketing Team and Product Team. Identify various factors and design key product/business metrics • Developed a generalized field extraction module which successfully extracts fields from complex scanned documents using machine learning and heuristic approach with >95 % extraction accuracy • Develop and support software including applications, database integration, interfaces, and new functionality enhancements. Create Data models and mathematical models of large data sets. Develop reports to make data meaningful to business initiatives, analyze data quantitatively and qualitatively and communicate findings. •Answering technical requirements for use cases and proof of concepts (POCs)

Clients: OKO, Precure, Acua Ocean, Omdena France,Archilyse • Collaborated to build custom semantic segmentation model for the analysis and monitoring the efficiency of Automatic Waste Segregator system using MASK RCNN • Led a team to Build an AI solution that identifies and classifies illegally operating vessels and extracting text information using pre-trained EasyOCR in partnership with Acua Ocean • Developed custom models using ensemble learning algorithms to predict painful behavior of tennis elbow using the muscle and movement data from MLI® Wearables in partnership with Precure

• Working in two funded projects, various internal projects and guided three Ph. D candidates. Machine learning model developed for basic classification. Built custom deep learning models to predict the depression level using speech and further work is under progress. Botnet detection in IOT networks using Machine Learning • Manage databases and home-grown tools and identify ways to use existing data in more productive ways. Possess knowledge and experience in the acquisition, analysis, and display of data and in extraction of information and knowledge from that data in supports decision making to business leaders. • Work on process efficiency by automation and smart analytics to remove redundant activities. Provide ongoing analytical support to Business Division and various levels of management in the organization related to sales improvement and other issues. Initiate discussion of projects with stakeholder and drive the same. • Built Multi-class classification model for classification of cancer cells based on light intensity • Launch new products and features, test their performance, and iterate constantly and quickly. Identify opportunities and define product vision and strategy. Automated assessment of psychiatric disorders using speech collaboration with Cooper Hospital, Vile Parle west, Mumbai (Ongoing). Co-Investigator • Data creation using NetSim-Simulator for creating various botnet attacks in IOT devices. Built imbalanced Classification model for Botnet Detection in IOT and further work is under progress • Automated assessment of psychiatric disorders using Machine Learning

➺ 𝐂𝐨𝐫𝐩𝐨𝐫𝐚𝐭𝐞 𝐓𝐫𝐚𝐢𝐧𝐞𝐫: Data Scientist Course (EDA, SVM) ➺ 𝐂𝐨𝐧𝐭𝐞𝐧𝐭 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫: EDA and Supervised Learning (Regression)

𝘔𝘶𝘭𝘵𝘪-𝘥𝘪𝘴𝘤𝘪𝘱𝘭𝘪𝘯𝘢𝘳𝘺 𝘌𝘯𝘨𝘪𝘯𝘦𝘦𝘳𝘪𝘯𝘨 𝘤𝘰𝘭𝘭𝘦𝘨𝘦 𝘸𝘪𝘵𝘩 1400 𝘴𝘵𝘶𝘥𝘦𝘯𝘵𝘴 𝘢𝘯𝘥 120 𝘢𝘤𝘢𝘥𝘦𝘮𝘪𝘤 𝘴𝘵𝘢𝘧𝘧 ➺ 𝐋𝐞𝐜𝐭𝐮𝐫𝐞𝐫 and 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐆𝐮𝐢𝐝𝐞 for B-Tech Biotechnology students ➺ Active member of exam cell, R & D cell, progress assessment committee ➺ Lead for setting fluid flow labs

• Worked for two funded projects and various internal projects • Assist program director and senior operations and academics managers in planning on-campus sessions, preparing schedules and grading. Work with the HOD in creating and executing a training program. • Maintain the skills matrix for the department. Run the onboarding training program for new artists. Source Code Plagiarism Detection Engine (Department of Information Technology funded) as Co-Investigator • Creation of Linguistic Tools and Corpora for Indian Languages (Ministry of Human Resource Development Funded). Analysing and Reporting, Presentation, Financial Budgeting. Database preparation, Tool testing.