Optimized RAG and AI agent performance using prompt engineering, embedding tuning, and document chunking to improve chatbot accuracy and reduce API costs. Developed a Claude multi-agent system using customized MCP servers to automate sales reporting, customer inquiries, and document processing. Implemented multi-layered LLM guardrails for input/output to protect against prompt injection, jailbreaks, and data exfiltration attacks.

Prompt Engineer at Meta (via TEKsystems)

June 1, 2025 - January 1, 2026

Crafted and evaluated high-quality prompts to enhance performance of Meta’s LLM for Instagram search. Annotated and structured large-scale text datasets to support RLHF for LLM optimization. Identified and documented hallucinations, biases, unsafe content, and policy infractions for escalated engineering action.

AI-Driven Real Estate Data Specialist at Jovi Realty

February 1, 2025 - May 1, 2025

Developed and deployed RAG-based chatbots using LangChain, OpenAI, MCP protocols, and vector databases for intelligent, context-aware Q&A from proprietary real estate data. Built AI agent workflows using CrewAI, AutoGen, n8n, and OpenAI Agent to streamline lead management and follow-ups.

AI Engineer at RECKLI

July 1, 2021 - February 1, 2025

Built and deployed production ML systems on AWS (SageMaker, Glue, S3, Lake Formation) with MLflow tracking and model monitoring for drift detection supporting forecasting, churn, and anomaly detection. Productionized ML services with FastAPI, Docker, and CI/CD on AWS to reduce undetected errors by 70%. Secured LLM inference endpoints using AWS native security controls (Cognito OAuth 2.0, API Gateway authorizers, WAF rate limiting). Developed LLM-based call analysis pipelines to transcribe sales calls and extract objections, interest signals, and follow-up actions for CRM; also built estimation models for project cost, materials, and lead time prediction.