Hi I'm Joshua. I architect and build the "glue code" that connections cutting edge llm and RAG tools with high-scale data infrastucture. My core focus is reducing developer friction by creating robust, observable, and performant extensions for modern MLOps frameworks

I'm passionate about building the custom solutions and integrations that bring advanced AI out of the lab and into production for user-focused use cases.

• Starburst Contributiosn & Field Enablement:
  • My work centers on developer and operator efficiency.
    • Contributions to Minitrino (lightweight testing envirionments)
    • Comprehensive Trino Runbooks for advanced debugging
    • The Terraform toolkit for declarative data infrastucture deployment

Expertise:

• This experiences has made me an expert in deploying, optimizing and troubleshooting complex Distributed Systems and Kubernetes/Container infrastucture for petabyte-scale data access.

• I'v been applying my infrastructure knowledge to the open source AI landscape, specifically by customizing and extending foundational tools to handle enterprise data complexity
  • RAG & LLM Tooling:
    • I build direct contributions and glue-code customizations to RAG libraries like LangChain and LlamaINdex, ensuring they can robustly connect to and query structured data sources like Trino
  • MLOps:
    • I create specialized extensions on modern MLOps frameworks like Truss and Opik to solve production challenges like structured output, observability, and efficient model serving

• I'v been applying my Kubernetes experties to adding Deployment guides for deploying performant Trino Clusters on Kubernetes
  • Trino-Gateway-API
    • https://github.com/joshuaFordyce/trino-gateway-api
    • This repository provides the Kubernetes Gateway API resource definitions to expose the internal Trino Coordinator Service externally
  • Trino-K8s-Cilium-Perf
    • https://github.com/joshuaFordyce/trino-k8s-cilium-perf
    • This project provides the configuration blueprint and deployment scripts to establish a high-performance network foundation for Trino running on k8s, leveraging Cilium with eBPF

These are currently in-progress or recently completed projects that demonstrate my focus on field-ready AI solutions:

• Semantrino

  • https://github.com/joshuaFordyce/Semantrino
  • Provides a tested, production-ready bluperint for high-accuracy RAG on SQL data

• Truss VLM

  • https://github.com/joshuaFordyce/Truss-Example-Rag-Agent
  • The TrussVLM template solves the core challenge of deploying sophisticated Vision-Language Models ($\text{VLM}$s) by creating a secure, robust I/O Gateway directly within the Truss Python environment.

• VectorHub

  • https://github.com/joshuaFordyce/VectorHub
  • Provides a vector store agnostic federalized client for retrieving vectorized data from multiple Vector Stores at once

• LLamaIndex Trino Reader

  • https://github.com/joshuaFordyce/llama_index
  • Improves efficiency and integration for data loading into vector/indexing tools

I am actively seeking roles like forward deployed engineer, field deployment engineer and Post Sales Solution Architect where I can directly bridge the gap between cutting-edge AI research and real-world customer challenges

• linkedin: https://www.linkedin.com/in/joshua-fordyce-197b89186/
• Personal Blog: https://infrastructurecorner.com/