NICK POTTER
All work
In Production2025 — present· Employer product · paiv.ai

Paiv

Production AI agent that validates insurance-claim estimates in real time.

At Paiv I built and own the backend powering a text + voice AI agent that catches missing photos, details, and documents before insurance estimates reach third-party reviewers — and set the team's standard for how agents are engineered.

~100k
events / day
Text + voice
agent surface
Real time
estimate validation

Paiv puts an AI agent in the loop of the insurance-claim estimate process. I own the backend that makes the agent reliable enough to trust with real claims — the tool layer, the data queries, and the orchestration — and I've become the person the team relies on for how to build with coding agents.

The agent

Catching problems before a human ever sees them

The agent — over both text and voice — validates claim estimates in real time, flagging missing photos, details, and documents before they reach third-party reviewers. I engineered the tool layer, database queries, and orchestration that make those judgments accurate and fast.

Agentic engineering

Setting the team standard for building with agents

I designed custom skills, hooks, and steering files for Cursor, Claude Code, and Kiro that let coding agents autonomously query DynamoDB and RDS, deploy through SAM, and parse CloudWatch logs. It became the resource other engineers use to set up their own agent workflows.

Reliability at scale

Data integrity across ~100k events a day

I'm trusted with reliability and data-integrity work across an event-driven pipeline handling roughly 100,000 events daily — diagnosing concurrency bugs, hardening services, and protecting key customer accounts. When the team contracted, I took on broader ownership across the whole stack.

Full stack

Whichever layer needs to move

Agent backend, real-time messaging on Lambda + AppSync + DynamoDB, React and React Native frontends — I pick up the layer the work needs. As the team got leaner I reduced tech debt by consolidating feature flags and configuration into org-level settings, sunsetting low-usage features, and pruning backend services.

Related writing

Next project

Prior