Case Studies
Case Studies
I came to this work from fifteen years of managing client relationships, and I learned the engineering by shipping it. I have rebuilt voice AI inside a live operation running thousands of calls a day, advised product teams on where AI belonged in their roadmap and where it did not, and taken a production system of my own from the first architecture decision through deployment. Different businesses, different scopes, and the same job underneath: the technology performs when one person owns it. The engagements below show what that looks like.
Engagement 01 · Embedded Hands-On
Boost Health Insurance
Boost runs 10,000 to 20,000 outbound calls a day. An enthusiastic, semi-technical COO had brought in voice AI and gotten stuck where most companies do: a great-sounding web demo that fell apart on real phone calls. Latency was over a second, callers were hanging up, and the focus had drifted to how the voice sounded instead of how the calls performed.
I started by verifying their numbers, and found that 67% of the transfers their platform reported as successful were actually voicemails. No one had caught it. From there I fixed what the data showed.
- Rebuilt voicemail detection, then migrated platforms for answering-machine detection at the telephony layer
- A self-improving agentic dialer: an AI agent writes challenger scripts from call outcomes, tests them head to head with A/B testing, and shifts call volume to the winner on its own
- An Asterisk SIP bridge connecting modern voice AI to their legacy ViciDial dialer, passing each caller's number through intact so their system routes it correctly
- Live reporting and post-call AI classification: human, voicemail, or robo-screener, on every call
- A HIPAA agreement with the platform vendor that is normally reserved for $160K enterprise contracts
Public Upwork review
"We ended up doubling our transfer rate ahead of target... He communicates clearly, gives helpful context around any issue that comes up, responds quickly, and stays solutions-focused."
Boost Health Insurance
Engagement 02 · Advisory
Foundry
Foundry is a job management platform for home-service contractors: HVAC, plumbing, electrical, the trades. The team brought me in to scope where AI added real value and where simpler rules-based structure was the better call.
The deliverable was an AI integration brief: which features deserved a model, which needed deterministic logic, and how to sequence the MVP so AI risk never blocked shipping.
Public Upwork review
"Cameron was exceptionally effective at helping us scope AI correctly, where it adds real value and where rules-based structure is the better solution. His guidance reduced build risk, clarified MVP priorities, and helped our team move faster with more confidence. He's sharp, pragmatic, and focused on shipping cleanly rather than over-engineering."
Eric Kaplan, CEO, Foundry
Engagement 03 · In-House Product
Bela Helps
Bela Helps is an after-hours voice AI for service businesses. Bela answers, books, escalates, and hands you the message. I serve as the fractional CTO, and I have made every technical decision on the product, from architecture through deployment.
- A custom pipeline on Google Cloud: agents on Cloud Run, self-hosted speech models on GPU, Twilio telephony
- A deterministic booking flow. The model physically cannot hallucinate a booking; only the booking step holds the booking tools
- HIPAA-compliant architecture
- Multi-tenant from day one: one platform, many businesses
The product takes live calls at (423) 734-9401, and you are welcome to call it and judge the work directly.
Book A Working Session
Where does your
voice AI stand?
Tell me about your voice AI and where it is stuck.
One hour, and you will know where your build stands and what to do next.