10+ years building large-scale data infrastructure and cloud platforms across healthcare, fintech, banking, and insurance — with deep expertise in Amazon Connect, PCI-DSS/HIPAA compliance, and AI/ML integration.
Takeaways from Philipp Schmid’s talk — write 5 real cases, strip no-ops, run ablation tests, and 10 best practices for evaluating agent skills (plus how DeepMind gates skill PRs on measured lift).
Authentication vs. deepfake detection — how voice-clone fraud actually happens, the scale of the losses, the five-layer detection stack, and how contact center platforms are responding.
Recognizers slice audio into fingerprints, match them against learned prototypes by dot product, and drift into errors when reality moves off the manifold — while softmax keeps them sounding sure.
8 VC firms, billions of dollars, one category. What the capital flows reveal about where voice AI value will land — and which layers nobody is funding yet.
A layer-by-layer breakdown of every major player — foundation models, real-time infra, vertical agents, and QA — what they actually do, how they differentiate, and who’s winning.
Why voice AI is the biggest platform shift since mobile — what changed in 2024, the four-layer stack, and why “just add voice” is a trap.
Senior Staff Engineer with 10+ years building large-scale data infrastructure across healthcare, fintech, banking, and insurance. I specialize in cloud architecture, ETL systems, and contact center platforms — with a focus on delivering outcomes that actually move the needle, not just shipping code.
Beyond the technical work, I'm invested in growing engineering teams — mentoring engineers, raising the bar on architectural thinking, and helping organizations make better decisions with their data. Research and continuous learning are non-negotiable parts of how I operate.
This site is where I write about the work: deep dives into AWS and multi-cloud architecture, IaC patterns, data engineering at scale, and whatever I'm actively researching or building. If something took me time to figure out, it probably belongs here.