Rishavraj Mandal

- Launched an AI Copilot for IFRS 9 (SPPI) classification, enabling analysts to automate financial instrument review, improving throughput by 38% across 10,000+ documents. - Engineered production-grade FastAPI microservices for metadata management, integrating Azure Blob Storage, Azure Table, and Redis to support high-scale parallel processing. - Orchestrated prompt chaining using Azure OpenAI and Azure Cognitive Search, enabling semantic Q&A on legal prospectus documents; cut manual data lookup time by 45%

- Deployed and fine-tuned using OCR and BERT-style sequence tagging, improving attribute accuracy by 24.6% across large retail catalogs. - Optimized a large-scale data ingestion pipeline using Azure Data Lake Storage and Apache Spark, processing 20TB of retail product data and storing enriched ASIN - Scaled end-to-end ML training by implementing Jenkins-based CI/CD pipelines across 8+ NVIDIA A100 GPU nodes via Kubernetes orchestration, reducing average training duration by 75% (from 14.2 hrs → 3.5 hrs)

Implemented a time-series model for demand forecasting, curated massive datasets from various sources, and achieved precise demand identification for insights. Conducted A/B testing and quality assurance by identifying optimization opportunities, leading to a 15% boost in overall model accuracy and user satisfaction. Implemented unit tests with a coverage of over 90% to thoroughly validate the time-series demand forecasting model's functionality and accuracy.

A project to create a responsive gaming website based on the UX design using HTML, SCSS, and React-JS. •One can find the latest articles and reviews on newly released games.