_

Vismay Devjee

Engineering Intelligence for Finance. Specializing in RAG workflows, LLM evaluation, and Anomaly Detection systems for high-stakes environments.

View Deployed Modules Experience Logs Download CV

Experience Logs

Fidelity Investments
Risk Analyst (Jan 2025 - Present)
Boston, MA

Driving SOX certification controls and implementing AI-driven risk oversight.

System Architecture: Expense Ratio Anomaly Detection
Fund PDFs
Python/Textract
Stat Models
KRI Generation

System Architecture: Financial RAG Pipeline
Financial Query
Vector DB Retrieval
LlamaIndex Eval
Reasoning Model (Gemini/Claude)
  • Manage SOX certification controls ensuring alignment with enterprise risk frameworks.
  • Developed LLM RAG applications integrating fund documents into vector databases.
  • Engineered custom LLM evaluation frameworks using LlamaIndex.
Deloitte LLP
Data Engineer (Oct 2022 - Aug 2023)
Mumbai, India
  • Led multi-phase AI chatbot development for Amazon using RLHF and Python.
  • Implemented robust categorization systems with statistical modeling to categorize tone and context.
  • Supported Enterprise Risk Management initiatives across technology infrastructure.
Stay Vista
Data Analyst Intern (Feb 2022 - Oct 2022)
Mumbai, India
  • Designed product to analyze response time queries, reducing agent response time by 20%.
  • Visualized booking queries to support informed decision-making and product strategy.

Deployed Modules

Expense Anomaly Detector

Automated system extracting critical fields from Fund PDFs to generate Key Risk Indicators (KRIs) for non-compliant funds.

Python AWS Textract Stat Modeling

Financial RAG Pipeline

Retrieval-Augmented-Generation workflow for financial queries. Includes custom evaluation frameworks to assess reasoning models.

LlamaIndex Vector DB Gemini/Claude

RLHF Chatbot Engine

AI chatbot development emphasizing tone categorization and context awareness through Reinforcement Learning from Human Feedback.

Python RLHF NLP

Response Time Analyzer

Analytical tool deployed to reduce agent response times by 20% through query visualization and pattern recognition.

Data Viz Stat Modeling Python

Technical Stack

Generative AI & ML

LlamaIndex LangChain RAG Vector Database OpenAI API Gemini Claude Prompt Engineering RLHF NLP

Cloud & DevOps

AWS Textract Azure Docker Kubernetes CI/CD GitLab Apache Airflow

Data Engineering

Python SQL SparkSQL Snowflake MongoDB FastAPI Statistical Modeling