Bhavya Soni

I have worked with Crest Data for 3+ years as a software engineer.

Project: Netskope Cloud Exchange

Description:

• Played a key role in the development of the Netskope Cloud Exchange platform.
• Designed and implemented a highly efficient data retrieval mechanism with data transformation and ingestion, enhancing the capability to extract extensive log volumes from the Netskope Tenant to the Netskope Cloud Exchange platform. Optimized the performance of the platform by remarkable 50% increase in throughput while reducing RAM usage by 2x.
• Built a circuit breaker mechanism to ensure stable data ingestion under heavy loads, improving system reliability and reducing downtime.
• Developed Netskope loT module to transfer the high volume of assets data from 3rd party platforms to the Netskope Device Intelligence platform with supported transformation.
• Managed Python async tasks using Celery and RabbitMQ broker.
• Technologies used: Python, FastAPI, MongoDB, Pydantic, Celery, Docker, RabbitMQ.

Project: HRMS -Internal Tool

Description:

• Handled a management portal in Python, Flask, PostgreSQL, Docker.
• Collaborated with cross-functional teams to deliver critical enhancements and bug-fixes.
• Redesigned an existing feature using hashing with reduced time-complexity of 0(1), at the expense of extra space.

Project: Netskope AI plugin

Description:

• Automated common query tasks by implementing a Netskope Al plugin with Amazon Bedrock, enhancing organizational efficiency by enabling rapid, streamlined access to Alerts and Events data within the Netskope Tenant.
• Developed an AI agent that intelligently redirected queries to appropriate tools, facilitating data extraction from plattorms such as Netskope, Crowdstrike, and Okta.
• Integrated documentation support by enabling the chatbot to answer queries related to Netskope Cloud Exchange, providing users with quick access to relevant information.

Also I have hands-on experience with generative AI. I have built some chatbots and AI tools for my current organization. And also delivered the third best solution of Gemini Ultra 1.0 hackathon by Lablab.ai platform.

Personal Projects:

Data Cleaning and Querying Engine

This project leverages advanced Large Language Models (LLMs) and supporting frameworks - Langchain Pandas Agent to clean, transform, and analyze flight booking datasets. The goal is to empower users to handle inconsistencies, fill missing values, rename columns for business-friendly querying, and perform data-driven

analysis through an intuitive interface.

Chatbot for API Documentations

Integrated Notion APIs to extract data from Notion pages in markdown format and efficiently saved it into text files for further processing.

Processed and indexed large datasets by creating chunks using MarkdownHeaderTextSplitter and

RecursiveCharacterTextSplitter, generated embeddings with Cohere's embed-english-v3.0 model, and stored them in ChromaDB.

Implemented a Reliable RAG (Retrieval-Augmented Generation) pipeline by leveraging LLMs to validate context relevance of data retrieved from the vector database, ensuring output groundedness, and detecting hallucinations in LLM-generated responses.