I'm Jonathan. I've been building production systems for about 5 years, the last two on AI. Voice agents, intake automation, and the Go services they sit on. I run Brokerloop, an Australian insurtech based in Melbourne. Most of what I ship there ends up as voice intake, long-form workflows, and the kind of integrations brokers actually use, not the kind that sit in a deck.
I came into AI from full-stack engineering and consulting, which isn't the usual route — most come through research. Most of the work has been greenfield, from small startup builds through to larger production systems. That's where the commercial-first instinct comes from: what does it cost, what's the scope, what happens when it breaks. The model is one piece of the system, not the system.
This site is mostly notes I keep while shipping things. Voice AI bits, Go backend bits, and what it actually takes to run an insurtech where compliance and real customers both apply pressure at the same time.
Outside Brokerloop I take a small number of contracts each year, usually voice AI or backend infrastructure, usually for teams with a prototype that works in a demo but not in production. New work is by referral. Email's the easiest way: jabrahamtech@gmail.com.
Selected work
Mostly greenfield — small startup builds through to large production systems. Confidential where noted.
- Fraud detection AI · tier-one Australian bank. Architected an AI fraud-detection system inside one of the big four. Internal-facing, regulated environment, lived inside the bank's stack with all the audit trail that implies. Confidential client.
- US engineering contract · funded LA startup (2024). Helped a Los Angeles company ship their first venture-backed product. Shipped engineering across the time gap. Confidential. Open to more US work.
- Brokerloop · founder. I run this. Voice intake for Australian insurance brokers, plus the Go services underneath — integrations with broker platforms, long-form workflows, and the parts that have to keep working when call volume spikes.
- IoT automation · largest laundry in the southern hemisphere. Built the automation layer connecting industrial machines to operations. Real machines on a real floor, throughput at scale, runs hands-off. Range work — proves the engineering travels outside AI.
- Low-latency event-driven architecture. A sub-second decision pipeline I originally built around prediction markets. The architecture works anywhere reaction time is the bottleneck: trading signals, real-time risk, anything event-driven.
- Open source · what I'm reading. I read more code than I write publicly. Currently following nautilus_trader (Rust trading engine), ComfyUI (diffusion graph), and open-webui (local LLM tooling).
Stack
- runtime: go · typescript · python · react · postgres · redis
- ai: openai · anthropic · groq · langgraph · agent skills + plugins · mcps
- voice: twilio · deepgram · retell.ai · elevenlabs
- ops: aws · railway · vps · cloudflare · sentry · posthog
- onchain: base · solidity contracts · ponder.sh
Elsewhere
- email — jabrahamtech@gmail.com
- github — /jabrahamtech
- linkedin — /in/jabrahamtech