ASTHA JAGGI

Gen AI Solution for Hitachi

• Designed and implemented an end-to-end Quality Inspection Tool leveraging Python, Generative AI (Azure OpenAI LLMs for data extraction), and overlay techniques to automate quality checks on new and legacy drawings. The solution compares document templates against a predefined attribute checklist across multiple document types, streamlining verification of Product Brand Migration changes and reducing manual effort.
• Built with Windchill for sourcing data, Azure Blob Storage for secure data handling, OpenCV for image processing and overlay, and a Next.js frontend for user interaction.
• Provided an additional option for manual alignment in the overlay, enabling users to select four points and realign images for overlay whenever required.

Generative AI Capability build for ITC Infotech

• Synthetic Data Generation
• Validation and Summarization of ESG reports using GAN's and VAE's

Erste Digital

• Performed an extensive exploratory analysis on the Credit Lab Reports (ESG) data shared and created a dashboard to visualize the important insights.
• Built an interactive dashboard using Dash Plotly highlighting important graphs/charts which would be integrated later to CreditLab portal.
• Performed migration of codes to Bitbucket and code deployment using their inhouse tool ChopSuey.

Accor

• Created several proposals for GenAI use cases - HR Chatbot, IT-Ops chatbot, D&TS and GAIA onboarding.
• Built Gen AI solutions for Accor - Chatbot for HR operations, IT-Ops Operations Chatbot using Azure AI Studio, Azure AI services, and other Azure platforms with functional specifications such as Search and Retrieval, Chat with knowledge Base, Summary and Translation, Document Tagging.

UK Based Retail Pharmacy Chain: Built a 360-degree single view of patient. Key solutions involved:

• Defining churn and building a Patient Churn Propensity Model.
• Patient Lifetime Value Prediction
• Customer Cohort Analysis – Behavior based segmentation of patients

Banking product development for Client pitch

• Objective – To build Analytical Models - across key moments of truth of Retail customer journey and retain core customers by timely offers and marketing actions.
• Approach – Identifying the suitable use cases (Loan Delinquency, Customer Churn, Identification of AML False positive alerts), curating, and simulating the datasets according to the business requirements and developing machine learning models on Python for the identified use cases.
• Algorithms Used – Logistic Regression, Random Forest, XGBoost

For a leading CPG/Retail Company

• Fashion Retail:
• Objective – Build Advance Analytical Solutions for effective DM campaigns and communication strategies.
• Approach – Customer level segmentation and profiling was done initially to define the customer persona based on their historical behavior, need and attitude towards shopping. Various seasonal and festival based predictive models were then built on these segments like Visit Propensity Models (Black Friday, Season Launch), Churn Arrest Model, Repeat Purchase Propensity Model, Brand Propensity Model, Customer Acquisition Models etc. The results were then utilized for customer targeting.
• Algorithms Used – K-means, Logistic Regression, Random Forest
• Objective – To identify and target potential chunk of customers for a Brand-Gender combination who behave like our existing customers using like to like mapping.
• Approach – Look-Alike mapping is performed between two groups of customers having similar attributes by using distance metric (Euclidean distance) and the Brand-Gender combination with the minimum distance value is assigned against that customer. Hence a customer can be targeted for that Brand-Gender combination.
• Method Used – Euclidean Distance Metric

Foods Retail:

• Objective – Personalized Product Recommendations based on customer persona.
• Approach – Based on the customer persona, three types of recommendation engines were built for cross-selling/upselling of products. These included: New Product Recommendation, Repeat Purchase Recommendation and Complete the Basket Recommendation.
• Algorithms Used – Collaborative Filtering, Apriori, Random Forest, GLRM

For a leading Airlines Company:

• Objective – Food wastage and pricing analysis for food inventory restock, wastage minimization and cost reduction.
• Approach – The extensive analysis was done and a client report was made which involved the monthly trends for item wastage, item-wise total wastage percentage, weekly pattern for veg and non-veg wastage percent, uplift station wise wastage trend and top wastage routes. This analysis helped the Client to take necessary steps to reduce the wastage and limit the onboarding food.