Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Generative AI engineer focused on LLM/RAG systems, agent workflows, and evaluation/guardrails in production environments with strong attention to data privacy and CI/CD release quality. Built architectures using LangChain/LangGraph, vector and metadata storage, and fast, traceable inference pipelines. Previously developed client-facing AI services and production applications across web and mobile, including schema-validated tool calling and offline-first evaluation pipelines. Worked end-to-end from infrastructure and deployments to testing, monitoring, and performance/cost improvements.…Generative AI engineer focused on LLM/RAG systems, agent workflows, and evaluation/guardrails in production environments with strong attention to data privacy and CI/CD release quality. Built architectures using LangChain/LangGraph, vector and metadata storage, and fast, traceable inference pipelines. Previously developed client-facing AI services and production applications across web and mobile, including schema-validated tool calling and offline-first evaluation pipelines. Worked end-to-end from infrastructure and deployments to testing, monitoring, and performance/cost improvements.

Ryan Mitchell

AI Engineer, Software Engineer, AI Developer, +2





Generative AI engineer focused on LLM/RAG systems, agent workflows, and evaluation/guardrails in production environments with strong attention to data privacy and CI/CD release quality. Built architectures using LangChain/LangGraph, vector and metadata storage, and fast, traceable inference pipelines. Previously developed client-facing AI services and production applications across web and mobile, including schema-validated tool calling and offline-first evaluation pipelines. Worked end-to-end from infrastructure and deployments to testing, monitoring, and performance/cost improvements.…Generative AI engineer focused on LLM/RAG systems, agent workflows, and evaluation/guardrails in production environments with strong attention to data privacy and CI/CD release quality. Built architectures using LangChain/LangGraph, vector and metadata storage, and fast, traceable inference pipelines. Previously developed client-facing AI services and production applications across web and mobile, including schema-validated tool calling and offline-first evaluation pipelines. Worked end-to-end from infrastructure and deployments to testing, monitoring, and performance/cost improvements.

Available to hire

Generative AI engineer focused on LLM/RAG systems, agent workflows, and evaluation/guardrails in production environments with strong attention to data privacy and CI/CD release quality. Built architectures using LangChain/LangGraph, vector and metadata storage, and fast, traceable inference pipelines.

Previously developed client-facing AI services and production applications across web and mobile, including schema-validated tool calling and offline-first evaluation pipelines. Worked end-to-end from infrastructure and deployments to testing, monitoring, and performance/cost improvements.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Work Experience

AI Architect / Generative AI Engineer at Orlando / FL Te ch Affinity, Inc.

October 1, 2024 - June 2, 2026

Architected and implemented LLM graph-based generative AI systems (LangGraph, vector search and metadata filtering) on AWS Bedrock and Kubernetes with access controls designed to keep client data inside each client environment. Built RAG pipelines and agent orchestration in Python, LangChain, LangGraph, and FastAPI. Maintained graded evaluation sets and used them to gate releases in CI/CD. Implemented safety measures including guardrails/PII redaction and validation of model/tool arguments using schema constraints (e.g., Pydantic) to reduce prompt-injection risks. Developed model evaluation and tracing workflows (token usage, retrieved context, tool calls) and improved inference efficiency (reduced latency/cost via routing and smaller model execution). Delivered to multiple industries (airline, healthcare, publishing, and technology services).

Mobile Application Engineer at AllianceTek Inc.

June 1, 2018 - August 1, 2021

Developed client mobile applications (iOS in Swift and Android in Kotlin/Jetpack) and supported production release processes. Built intake and chunking pipelines in Python using LlamaIndex, Docker, and PostgreSQL for document processing (PDFs/HTML exports/scanned mail). Shipped a customer support assistant for the OpenAI API with function calling and typed tool schemas, replacing ad-hoc string parsing with Pydantic models and structured outputs. Retrofit conversational AI into production-ready web and mobile products by versioning FastAPI endpoints and streaming responses. Implemented offline-first evaluation with Ragas/LangSmith; batched embedding jobs and caching to improve cost/latency while maintaining answer quality. Participated in CI/CD and quality tooling (XCTest, Espresso, JUnit) and monitored crashes via App Center.

AI Engineer at ExOne / NexOne Technology?

May 31, 2018 - September 30, 2024

- Fine-tuned BERT and DistilBERT in PyTorch with Hugging Face Transformers to route incoming requests and sort documents. The keyword rules they replaced kept breaking on real mail. - Added spaCy NER pipelines that pulled policy numbers, dates, and medical terms out of unstructured correspondence. - Built a question-answering service over a client library of 40,000 documents with FastAPI, LangChain, and pgvector on OpenAI embeddings. Every answer linked back to its source passage. Delivery teams stopped hunting through the corpus by hand. - Wrote the intake and chunking pipeline in Python with LlamaIndex, Docker, and PostgreSQL. PDFs, HTML exports, and scanned mail came out as clean searchable passages, one corpus across the airline, healthcare, and publishing projects. - Shipped a customer support assistant on the OpenAI API with function calling and typed tool schemas, and replaced the hand-written string parsing behind it with Pydantic models and strict Structured Outputs. Far more responses came back machine-readable on the first try. - Retrofitted conversational AI into client web and mobile products already in production. Versioned FastAPI endpoints, streamed responses, no redesign of the existing screens. - Set up an offline eval harness ith Ragas and LangSmith that scored faithfulness, answer relevance, and context recall against a fixed question set. - Batched embedding jobs, cached the common lookups, and pushed short tasks to smaller models. Cost and latency came down and answer quality held. - Turned exploratory notebooks into tested Python packages with real interfaces.