Rudra Prasad Bhuyan

Delivered an AI interview platform end-to-end for a confidential client, covering LLM question generation, real-time audio processing, and cost-optimized inference.

Designed 12 Star Schema databases to generate up to 15 relevant, non-repetitive interview questions per request.

Reduced resume-analysis cost to ₹0.13 per student, enabling analysis for approximately 100 students at about ₹13 total.

Built an airline Talk-to-Data application that uses RAG to retrieve relevant tables, columns, and relationships from internal documentation.

Implemented hybrid retrieval with semantic search and BM25 to ground a Text-to-SQL model in relevant business context.

Designed a three-model workflow for question understanding, fine-tuned Text-to-SQL generation, and SQL-to-natural-language responses.

Created a 400-sample golden dataset to evaluate accuracy and reliability across the Talk-to-Data workflow.

Guided a 4-person team on revenue-pipeline optimization for a 60-110 GB aviation dataset.

Project I have working Fin4Rural (Indian Level) | Bihar Dashboard (State Level).

Developed Python and SQL ETL pipelines for 15 multi-level HCES datasets with 400+ features to streamline large-scale batch processing.

Analyzed 500+ socioeconomic and healthcare variables to identify trends in finance, healthcare, consumption, and savings behavior.

Automated data validation and duplicate-handling workflows using SQL procedures, reducing data inconsistencies by 20%.

Built reusable data preprocessing scripts in Python, reducing manual model-preparation effort and saving ~1 hour daily.

Designed scalable system architecture, PRD documentation, and 30+ analytical reports to standardize workflows and business planning.

Built interactive KPI dashboards across 50+ parameters and collaborated with stakeholders to deliver actionable insights and summaries.