Himanshu Pant

Remote

• Developed multi-agent Stateful RAG system for personalized legal guidance using LangGraph with Qdrant vector DB, achieving sub-2s query latency across 10K+ documents and 95% relevance accuracy
• Built FastAPI + Node.js backend with Docker supporting 100+ concurrent sessions with sub-200ms response time through optimized vector retrieval

Gurugram, India · Remote

• Drove 40% BI query performance improvement by architecting S3 data lake and Snowflake warehouse across 8 datasets supporting 500K+ daily users processing high-frequency insurance claim events
• Eliminated 90-minute production bottleneck by optimizing PySpark ETL pipelines on AWS Glue/MWAA processing 100M records/batch, achieving 66% runtime reduction via partition pruning and Spark query optimization
• Built production anomaly detection layer using Great Expectations across 15+ data sources with CloudWatch monitoring and MWAA retries, maintaining 100% SLAs

Gurugram, India · On-site

• Reduced user churn 20% by building production ML retention pipeline with 78% accuracy using Scikit-Learn and statistical modeling that identified behavioral anomalies in time-series user activity data, deployed automated Python/SQL re-engagement campaigns
• Built and owned core data ingestion layer — multi-source batch ETL pipelines ingesting from 10+ REST APIs and S3 using cloud data solutions, loading 500K+ daily records of football, soccer, and basketball data into MongoDB with sub-hourly refresh cadence, enabling near-real-time analytics for product and marketing teams with 99.9% data accuracy via cross-source validation and data profiling
• Engineered ML feature pipelines with SCD Type 1/2 dimensional modeling and automated feature engineering, reducing ML experiment cycle time 35% through optimized time-series data storage patterns

Bangalore, India · Remote

• Spearheaded migration of 1 Billion+ record oncology pipelines from legacy SAS to Apache Spark on Azure Databricks, achieving 86% reduction in batch processing time from 6+ hours to 50 minutes through broadcast joins and strategic partitioning
• Automated bi-weekly HCP targeting reports by building Python ETL pipeline querying AWS Athena, reducing manual effort from 4–5 hours to 15 minutes, delivering 95% time savings
• Delivered 99.5% data accuracy post-migration via comprehensive PySpark and SQL validation frameworks with statistical reconciliation

Gurugram, India · On-site

• Built and maintained 15 Power BI dashboards with advanced analytics for executive leadership tracking project costs, timelines, and profitability across $50M+ in annual construction projects
• Consolidated cost data from multiple enterprise sources including SQL Server, Oracle, and MySQL via SQL queries and stored procedures, automating monthly executive reporting and reducing manual effort by 20 hours/month