Avantika Joshi

• Overview - This project is part of the Research and development team at Infogain.
• 1. Multi-Agent AI System for Business Analysis
• - Built a multi-agent system using LangGraph, LangChain, and LangSmith that interacts with Azure Devops.
• - Agents collaborated to analyze business requirements, generate insights, and automate project documentation.
• 2. Fine-Tuning LLMs for Predictive Modeling
• - Fine-tuned llama from hugging face on structured data (CSV with category, subcategory, description, and component name) to predict missing components.
• - Used LoRA, QLoRA, for efficient model adaptation.
• 3. Retrieval-Augmented Generation with Local LLMs
• - Implemented a RAG-based Q&A system using Llama models on Ollama for offline query resolution.
• - Used FAISS for embeddings and vector search.
• 4. Code Optimization and Bug Fixing Copilot
• - Built efficient prompts using OpenAI Codex to analyze, optimize, and fix Databricks Spark/PySpark code.
• - Enhanced performance by suggesting memory-efficient transformations and parallel processing strategies.
• 5. Visual Workflow Generator from Natural Language
• - Developed an LLM-powered app that converts text descriptions into flowchart diagrams.
• - Used Mermaid.js for rendering visual workflows.
• 6. RAG based Chatbot
• - Developed an end-to-end chatbot to answer user queries using vector db

Overview: This is a venture of client into digital space. We were responsible for creating an ecosystem in Azure Cloud.

• Design, implement, and maintain data pipelines for data ingestion, processing, and transformation in Azure
• Consulted and advised clients on resolving source end data anomalies and finalizing the implementation strategy
• Conducted working session with client during UAT
• Contributed along with stake holders in building new data sources for data ingestion
• Ingested data from disparate sources including Google Analytics, Salesforce and SharePoint using ADF and python to create data views to be build using Power BI
• Developed Python script to automate mails instead of logic apps reducing the cost by 80%
• Worked along with reporting team in writing DAX query and dashboard development in Power BI
• Mentored the team in resolving the bugs created during development
• Finalized the forthcoming project developments along with client, and presented to the stakeholders

Overview: This is a data analytics project for a client, with varied sources for their sales data across geographies

• Worked on development of logic apps for ingesting data from SharePoint to azure blob
• Developed Azure wiki pages on implementation steps of azure purview for the team
• Worked on a POC using Spark in Python to distribute data processing on large datasets, improving performance by 60% in comparison to existing views
• Implemented a feature to reduce the number of datasets in ADF using parameters, reducing manual effort while deploying logic apps
• Worked on other major and minor enhancements involving changes in existing ADF, Synapse stored procedures
• This is a data analytics project for a client, with varied sources for their sales data across geographies

Overview: This is a data migration project where the data from the different legacy data sources of client is migrated to Azure

• Worked on initial setup of Azure environment along with the Client's platform team, developed linked service, triggers and pipeline design
• Developed the data ingestion pipelines which ingested and transformed data post business rules for loading in to Azure SQL DW
• Played key role in Sprint & Capacity planning of the team using Azure Boards
• Mentored resources of the team on ADF helping them design the new requirements
• Interacted with the client on understanding of the new framework, gave KT to cross teams on its implementation
• Developed implementation document for the framework and helped in solving any issues
• This is a data migration project where the data from the different legacy data sources of client is migrated to Azure

Overview: This is a data analytics project where we handle the transactional and analytics data on Oracle and Hadoop respectively

• Lead a code migration from Unix to Python, migrating the existing Unix scripts into Python
• Developed HTML scripts for sending over mail via Python scripts
• Developed Python script comparing two Hadoop environments ensuring they are consistent across objects
• Developed a Python script which raised triggers in case higher probability of missing SLA
• Worked on automating sending a list of Autosys jobs status
• Completed several RCAs accurately and implemented enhancements
• This is a data analytics project where we handle the transactional and analytics data on Oracle and Hadoop respectively