Australian Software Developer for Voice Platform

We are seeking an Australian-based software developer to enhance our ai voice and chat platform. The platform is already integrated into various businesses and requires further development to improve its functionality and speed. The candidate will have deep experience with ai and voice and chat, and be able to work independently to deliver high-quality solutions in a quick time frame. Have real qualifications and must provide supporting evidence. Being ‘good at Claude’ is not good enough for us. Open to overhaul with strict commission mile stones prior to implementation. This is the second time using this approach and the first didn’t work to well. Waited 3 days for a sub par audit telling me what I already knew. I need someone who will hit the ground running immediately and get this resolved. Bonus payments and incentives will be made for meeting tight timeframes and passing pre Commision checks which we can discuss - so I’m less Interested in 3-4 months quoted work. I want it done now and will pay accordingly if the product meets the brief. Long term work available for right fit, we are an established business and the main software dev is looking to step away. Sound like you? Let’s chat because I’m ready to go today. More info on us. What we do We run a self-hosted voice and chat platform deployed into live business environments, currently serving seven active clients with scope to scale quickly. The platform answers calls, understands the caller, draws on a client-specific knowledge base, and resolves enquiries. Every deployment also ships with a chatbot running off the same knowledge base, plus a web scraper we built to populate and maintain that knowledge. Everything is hosted in Australia, and we have firm data sovereignty requirements. That constrains what can be swapped or moved, so the architecture is not fully open to substitution. Current stack • Voice: ElevenLabs • Transcription: Deepgram • LLM: AWS, with Azure as fallback • Telephony: LiveKit trunk, with Telstra and Twilio as number providers • Chatbot and web scraper: built in-house, sharing the same knowledge base as the voice layer Current round-trip latency sits at roughly 1.2 seconds. Workable, but not where I want it. The standout capability The most sophisticated part of the system is the after-hours logic. Clients build their own rule sets driven by semantic understanding. When a caller presents a problem, the system routes it into our protocol engine, and based on the client’s rules it dispatches an automated outbound call followed by an SMS to close the loop with the end user. This works very well today and is a genuine differentiator. It needs to be preserved and ideally hardened, not rebuilt for its own sake. Recent work I have just completed a significant overhaul to enforce one clear owner per function. Previously, overlapping “helper” logic was overriding things and generating persistent bugs. The codebase is now in a much stronger position: when something happens, I can identify the responsible owner and the cause. That accountability structure is non-negotiable and should carry into whatever comes next. A note on history: the codebase began life in Claude Code and accumulated layers over time. I have stripped a lot of that back, so it should be cleaner than you might expect from that origin. I am open to a candid review of where it actually stands. What I need Three problems matter most. 1. Latency. I want to push the round trip below one second so the experience feels effectively instantaneous. 2. Conversational quality. Right now, by design, the system too often defaults to “I don’t have that, call me back later.” I want it to converse naturally about anything within reason and come back with meaningful information rather than fillers or dead ends. 3. Hosted in Australia while having premium quality 11 labs voice that’s currently in use. Maybe open to discussing an alternate voice but latency from voice doesn’t seem to be the issue. Engagement approach I am open to two paths and want your honest view on which is right: • Optimise and finesse the current build, or • Redesign from the ground up if you believe that delivers a materially better outcome. I am not precious about the existing code. I am precious about the outcome. What success looks like • Sub-one-second response time • Natural, useful conversation that returns real information, not filler • A clear ownership structure mirroring what I have already built, so future faults are easy to trace and fix • The protocol engine and after-hours rules working seamlessly and reliably If you are responding, tell me which path you would take and why, and where you would start.