The problem
Transcription, redaction, intent detection, summarization, inferred CSAT, QA, and Voice of Customer each have different quality and operating requirements.
Level AI · Domain-specific AI platform
The intelligence underneath the experience.
Seven CX-specific AI capabilities. Product decisions that balance model quality with the realities of production.
The short version
Complexity,
made useful.
What I led
I own productization, evaluation systems, and infrastructure economics across seven CX-specific AI capabilities—balancing quality, reliability, latency, cost, privacy, and explainability.
What changed
Improved speech-model brand accuracy from 71% to 95% and delivered five-figure annual infrastructure savings. The focus is model capability that works within enterprise product constraints.
Inside the experience
Different jobs. Different intelligence.
Explore an illustrated product story.
The examples use sample data.
LATITUDEPRODUCT EXPLORATION
Purpose-built intelligenceMany models.
Many models.
One higher standard.
Contact-center speech → a usable transcript
Hear the words. Keep the context.
“I called yesterday about my refund. I was told it would arrive today, but I still haven’t received it.”
The product work
Where I
focused.
Strategy, decisions, and the details
behind the experience.
Make quality a product responsibility.
Connect model evaluation to the experience customers actually receive.
What the quality scope includes
My ownership spans productization and evaluation systems across seven capabilities. Accuracy alone is not the whole product requirement: the scope also includes reliability, latency, privacy, and explainability. The reported 71% to 95% improvement specifically concerns speech-model brand accuracy.
Treat economics as part of the experience.
Balance the quality of an AI capability with the cost and reliability of running it.
The production tradeoffs
Infrastructure economics is part of my product remit alongside quality. This work delivered five-figure annual infrastructure savings.
Work across the intelligence and application layers.
Connect the underlying AI platform with customer-facing products.
How the scope connects
At Level AI, my strategy scope spans the governed AI platform and agentic applications. I partner with the CEO and CTO and lead work across application engineering, ML engineering, and design, connecting model capability with product experience, commercialization, and adoption.
The outcomes
71 → 95%
speech-model brand accuracy
7
CX-specific AI capabilities
5 figures
annual infrastructure savings