engineer · builder · security tinkerer
Software engineer exploring the edges of AI, full-stack systems, and security. I build production grade systems for the curiosity of it and to learn what the demo never shows.
— usually with a half-finished idea and too much coffee.
See what I'm building Download résumé (PDF)building
A voice agent that answers real phone calls and holds a natural conversation built to replace a rented platform (Retell) with a voice layer that lives inside my own Spring application. Twilio handles the telephony; Parley owns everything from the words onward: interpreting the caller, running the LLM, and deciding what to say or do, with direct access to the backend. An experiment in owning the latency and turn taking that managed platforms hide.
An event-driven system that turns a live inbox into a searchable, classified knowledge base. It subscribes to new mail through Microsoft Graph, classifies intent with a local LLM, validates the result, and stores everything for RAG-powered agents that can reason over the whole mailbox cloud-native on Azure, provisioned with infrastructure-as-code. A study in how much judgment you can safely hand to an autonomous system.
A live, collector owned trading card business Pokémon, sports, and Magic: The Gathering. I build the storefront: a discovery first Next.js site with catalog browse and product pages, its inventory synced live from the eBay Trading API behind a cached read path. It's also the real product my AI experiments plug into.
also explored
Real-time messaging and persistence apps built from the ground up to understand the architecture behind production collaboration tools — sockets, concurrency, and local data not tutorial follow-alongs.
what I'm into
Applied AI
Full-stack
Infrastructure
Integrations
Security
writing
Event-driven mail classification, local LLM intent extraction, and RAG on Azure.
Why wrong retrieved context is worse than no context — and how to evaluate retrieval separately from generation.
blog
A markdown-first blog for technical articles — built at deploy time so Draftsmith can publish on-domain without hand-editing HTML.
All writing · Blog · Cross-posts on Medium use a canonical URL back to tommarler.com.
about
I'm a software engineer who rarely stays in one lane long enough to stop being curious about the others. My background sits at an unusual intersection iOS, distributed cloud systems, and hands on security research and that overlap is intentional.
The most interesting problems live at the edges between disciplines. Understanding the security layer makes you a better architect; understanding the product layer makes you a better engineer. Right now most of my curiosity is pointed at applied AI what it takes to make an autonomous system reliable enough to trust.
faq
I'm working on Parley, a self-hosted voice agent where I own the LLM layer inside my own Spring app; IntelligentOldMan, an experiment in turning a live inbox into something agents can reason over; and PokePitchShop, a live storefront where a lot of my AI ideas get tested against real traffic.
Managed platforms hide latency, turn-taking, and backend access. I wanted to see what it feels like to own those decisions — Twilio handles the phone call; everything from the words onward lives in my application.
An event-driven system that classifies incoming mail with a local LLM and stores it for RAG-powered agents. The interesting part isn't the stack — it's how much judgment you can safely hand to an autonomous system.
The gap between a convincing demo and a system you'd trust in production: agent loops that know when to stop, RAG that retrieves the right context, and voice agents where you own latency end to end. I learn by building, not by collecting buzzwords.
Columbia, Missouri. My background spans iOS, distributed systems, and security research — I tend to follow problems to whatever lane they lead.
Email me at tom@tommarler.com, or find me on GitHub at github.com/tommy3531 and LinkedIn at linkedin.com/in/tom-marler. Always happy to talk to people building interesting things.