Prince Brahmenia

Insights Prism

• Objective: Analysis on tableau dashboard is complex, to enhance user experience and for quick analysis using natural language for better negotiation and procurement
• Build RAG based architecture application with features like sql generation, visualization plot, dataframe creation, ambiguity resolution about named entities, summarization of sql response.

Insights LDS

• Objective: Deployed in manufacturing plant, used to analyze the performance, issues of line, sector, department

Version 1

• Designed custom tools to calculate defect rate, trend, efficiency, target and passed to zero react agent
• Applied semi structured RAG architecture, used RetrievalQA chain to extract context in informative manner related to user query

Version 2

• Shift from csv files to database, from custom tools and agents to custom chains, applied RAG based architecture with various chain like sqlquery generation, entity extraction, answer summary, dataframe creation, query refinement

Document Classifier

• Generate VectoreStoreIndex on storage bucket contains more than 40k docs, used chat model for classification into one of the five categories, attains accuracy of 80%
• Automate the process with error handlings, storing the embeddings of each document to avoid re-creation of index in case of any failure

Confidence Score

• Objective: Score which helps planner to trust on the recommendation given by us on safety stocks
• Enhance the score by changing the calculation of parameter based on the history of dc material level
• Developed the script which automatically captures the variation and responsible parameters
• Triggered monthly docker runs for confidence score and variation calcultation