Sandipan D.

Key Skills

Software

No software listed

Top Subject Matter

Enterprise knowledge management using LLMs (RAG, summarization, translation/localization)

Top Data Types

Document

Video

Top Task Types

Text Summarization

Classification

Freelancer Overview

Cisco — Data Scientist (LLM summarization + multilingual translation pipeline; RAG-based Q&A; agentic systems). Brings 7+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal and Proprietary Tooling. Education includes Post Graduate Diploma in Data Science and AI, IIIT Delhi (2022) and Bachelor of Technology in Electrical Engineering, RCC Institute of Information Technology (2017). AI-training focus includes data types such as Document and Video and labeling workflows including Text Summarization, Classification, and Fine

Labeling Experience

Data Scientist - Cisco

VideoClassification

Designed and operationalized an LLM-powered summarization pipeline for 1.1M+ enterprise documents, producing structured metadata for searchable datasets. Built multilingual document translation workflows to localize knowledge bases and remove manual translation effort. Architected production RAG, agentic AI, and multimodal retrieval systems that improved retrieval latency, reduced inference costs, and accelerated decision cycles for business teams. • Built LLM summarization, translation, RAG, and prompt caching solutions for enterprise knowledge discovery • Deployed deep-research and competitive intelligence agent workflows to automate market analysis • Implemented multimodal retrieval and reasoning across 40K+ multilingual audio/video assets using FFmpeg, Whisper, CLIP, and vector search • Developed ETL and analytics components using Airflow, Oracle, Postgres, and anomaly detection in Python

2024 - Present

Cisco — Data Scientist (LLM summarization + multilingual translation pipeline; RAG-based Q&A; agentic systems)

DocumentText Summarization

Built and operationalized an LLM-powered enterprise document summarization pipeline that generates structured metadata (file paths, filenames, summaries) into searchable datasets. The work involved preparing and processing large document corpora to support downstream retrieval and knowledge indexing for enterprise users. A key focus was producing training/knowledge artifacts (summaries and localized content) at large scale to improve retrieval quality and reduce processing latency. • Created structured summary outputs for 1.1M+ enterprise documents • Automated metadata generation for faster enterprise search and reduced turnaround • Enabled prompt caching and model usage optimizations for production RAG workflows • Supported multilingual localization via translation pipelines across multiple languages

2024 - Present

Business Intelligence Engineer - Amazon.com

VideoClassification

Built and deployed machine learning models in Python to support fraud detection and vendor quality analysis, improving audit accuracy and operational risk mitigation. Performed funnel, cohort, and root-cause analyses for product launches and developed time-series forecasting and anomaly detection pipelines for demand planning. Automated KPI reporting and dashboard creation using SQL and Python, driving faster executive insights. • Trained classification and regression models with scikit-learn to improve audit accuracy by 25% • Conducted cohort, funnel, and root-cause analytics to reduce return rates and improve retention • Implemented forecasting and anomaly detection workflows to raise forecast precision and supply chain planning accuracy • Built automated KPI reporting and 10+ dashboards in Tableau and QuickSight to reduce manual reporting effort

2020 - 2024

Education

Work History