Yashika Sharma

Leading development of “Tarko,” an LLM-powered chatbot using Azure AI Foundry and multi-agent orchestration to automate user support and streamline information access. Built scalable batch and streaming ETL pipelines using S3, PySpark, Delta Lake, and Kafka for financial data, reducing latency and enabling near real-time insights via BigQuery and Looker. Deployed intelligent search on Azure with OpenSearch, AKS, and hybrid reranking, improving retrieval quality and system scalability. Tuned OpenSearch ingestion pipelines and joint indices for faster indexing and better performance in production. Automated data refresh workflows with Azure Logic Apps and supported Power BI reporting, reducing manual work and improving data timeliness.

Developed a data governance engine with manual and automated tagging, supporting business glossary integration across UI, CLI, and API layers. Built and released a CLI (PyPI) using Typer and FastAPI, used by top 3 enterprise customers to manage metadata, trigger dbt actions, and surface regression reports on PRs. Improved impact analysis and lineage computation, reducing execution time by 30% and enabling column-level visibility. Extended the automated data catalog with manual lineage features to improve support for custom pipelines. Created onboarding and admin tools to streamline customer setup, reducing handoff time for sales and support teams.

Built an automated pipeline with Django and PostgreSQL to capture and store daily log data with sub-second latency. Collected and normalized data from 95K+ blockchain projects across CoinMarketCap and V.Cent.co for internal analysis. Integrated Typeform API to aggregate multi-form responses into AWS RDS, improving data consolidation and accessibility. Developed an Auth0-based authentication system to capture user metadata and support a proof-of-concept for Git-linked NFTs.