Jefferson – LLM, LangChain, RAG

Enterprise agentic AI platform for a global media network — multi-agent orchestration, production search and knowledge-retrieval applications, autonomous workflow automation, and cloud-native LLM pipelines with evaluation and observability. Covers AI application migration across cloud environments and end-to-end system delivery from architecture to production.

• Architected and deployed enterprise agentic AI systems using LangGraph, LangChain, CrewAI, and AWS Bedrock AgentCore, including multi-layer agent orchestration with controlled tool access between sub-agents.
• Migrated AI applications across cloud environments and built production RAG-based search and knowledge-retrieval applications.
• Designed context engineering strategies and governed LLM pipelines with structured evaluation, output quality measurement, and observability layers.
• Implemented cloud-native delivery on AWS with Terraform/Terragrunt infrastructure-as-code; integrated OpenAI API and Gemini API into production workflows.

Enterprise AI infrastructure for a programmatic advertising and data-analytics company — RAG platforms for large-scale knowledge retrieval over advertising campaign corpora, LLM evaluation frameworks for quality and grounding measurement, and vector-search pipelines supporting production retrieval at scale.

• Built enterprise RAG platforms and vector-search pipelines for document and advertising campaign corpus retrieval using OpenAI APIs on GCP.
• Designed LLM evaluation frameworks to measure output quality, grounding accuracy, and detect regressions across retrieval and generation workflows.
• Applied context engineering and prompt optimization techniques to improve retrieval relevance over large corpora.
• Built web-scraping and data-ingestion pipelines for knowledge-base construction supporting search and generation products.

Multimodal AI agent platform for SMB sales automation — combining voice AI (VAPI, ElevenLabs), telephony (Twilio), and CRM automation into autonomous outreach and conversation workflows for real-time business interactions and lead qualification.

• Built voice-first AI agents integrating VAPI, ElevenLabs, and Twilio API for autonomous phone-call handling, SMS workflows, and call routing.
• Designed multimodal orchestration pipelines connecting STT/TTS, telephony actions, and CRM updates via Make automation.
• Implemented conversational agent logic with OpenAI API for real-time lead qualification, call management, and follow-up workflows.
• Shipped end-to-end agentic automation replacing manual outreach and follow-up processes for SMB clients.

Healthcare AI platform for clinical text intelligence — transformer-based NLP pipelines enabling clinical and operations teams to extract structured information from unstructured medical documents, with PII/PHI masking for privacy compliance.

• Applied transformer models (Mistral LLM) to medical document intelligence for classification, named-entity recognition, and structured information extraction.
• Built healthcare NLP pipelines for clinical text processing covering NER, relation extraction, and production-grade retrieval.
• Implemented PII/PHI data masking and privacy-safe handling protocols for clinical data workflows.
• Delivered production-oriented NLP services integrated with clinical operations tooling.

Enterprise document intelligence and knowledge-management platform for a global manufacturing and technology corporation — semantic search, NLP-powered extraction, and retrieval systems over large technical and operational document corpora serving multiple internal teams.

• Designed and built semantic-search and NLP systems for enterprise document intelligence using Python, Azure OpenAI, and Microsoft Azure.
• Built knowledge-retrieval pipelines over technical corpora with LLM integration for structured extraction, classification, and summarization.
• Led a traffic-light quality evaluation system with a 2–3 person review team, measuring output quality and managing model regressions.
• Improved retrieval relevance, extraction accuracy, and document-understanding quality across classification and knowledge-access workflows.

Conversational AI and a semantic-search platform for a large B2B e-commerce infrastructure provider in Latin America, supporting catalog retrieval and customer-interaction automation.

• Built conversational AI and semantic-search systems over commerce catalogs using ElasticSearch and Python.
• Automated customer-interaction workflows with NLP classification and extraction pipelines for document and operational knowledge.