Applied AI · LLM Systems · Software Delivery

Daniel Aldron

I ship production AI features end-to-end — practical delivery over research papers.

Profile

Applied AI engineer who takes LLM products from idea to live, reliable and secure — and owns every layer in between. I build retrieval-augmented (RAG) chatbots and AI tooling on Azure: integrating OpenAI, wiring up vector search, hardening against real-world abuse, and running the deployment and monitoring myself.

Behind the AI work sits over a decade in software development and third-line support — Java microservices and cloud backends built inside large engineering teams as well as on solo projects, plus the kind of production firefighting that teaches you to build things that don't fall over.

Underneath all of it, decades of hands-on systems administration: Linux and Windows Server, self-hosted services, networking, storage and deployment pipelines. I run a substantial home-lab and a fleet of live sites and services on cloud infrastructure — so I'm as comfortable with the servers and the plumbing as with the application sitting on top.

What I build with

Applied AI / LLM

OpenAI GPT-4.1RAG Embeddings & vector searchPrompt engineering Prompt-injection defenceConfidence scoring

Backend & Cloud

Node.jsAzure Functions Blob StorageApp Insights REST APIsWebSocketsServerless

Integrations

Zoho SalesIQ / Desk / CRMStripe Hoowla APIPDF parsing & generation

Also in the toolbox

JavaPythonPowerShell GitBashHTML / CSS / vanilla JS

Systems & Infrastructure — decades hands-on

Linux server adminWindows Server Self-hosted servicesDocker nginx / reverse proxyDNS & TLS (certbot) WireGuard VPNZFS storage DigitalOcean / VPSsystemd · cron automation Networking & firewallsBackup & monitoring

Experience

AI & Technical Engineer — Dezrez
Sep 2025 – Present
  • Built and run “Dai”, a live customer-support chatbot (Zoho SalesIQ): RAG retrieval over a Zoho Desk knowledge base, OpenAI-generated answers with confidence gating and clean human handoff.
  • Hardened it for production — prompt-injection defences, internal-KB filtering, cross-org access control, PII stripping and history purge.
  • Shipped supporting tools as lean Azure Functions: time logging, Stripe reporting, document parsers, CRM data tooling. Own the architecture, deployment, security and monitoring.
Software Developer — Admiral (EUI Ltd)
2015 – 2018 · dev & e-learning roles through 2021
  • Developed business-critical claims applications, including Java microservice architectures; third-line support and performance optimisation.
  • Built cross-platform e-learning modules and maintained training-system infrastructure; vendor liaison.
Company Director — Way Out Pictures Ltd  ·  VFX Artist (freelance)
2019 – 2025
  • Ran an independent film company and freelanced in VFX — while owning all IT infrastructure, deployment, security and high-performance compute / digital-asset pipelines.

What sets the work apart

End-to-end ownership

Idea → architecture → build → deploy → monitor. I can own the whole lifecycle of a production AI feature, or slot into a team on any part of it.

Security-minded

PII, prompt-injection and access control treated as first-class from day one — not bolted on later.

Pragmatic engineering

Low-dependency, self-contained builds. Fewer moving parts means less to break and more to ship.

App to bare metal

Comfortable the whole way down — from the application code to the servers, storage and networking it runs on.