Architecting production voice AI agents using multi-agent orchestration and NLP-driven conversation flows for insurance FNOL (First Notice of Loss). Designed an agent architecture workflow that reduced voice agent build time from 8-10 days to under 1 day, a 10x acceleration adopted across the engineering team. Built reusable AI agent skills, prompt templates, and diagnostic tooling that enabled the broader team to ship agents faster. Delivering end-to-end voice flows for 5+ insurance clients, integrating with policy management systems like Guidewire.

Led end-to-end integrations between Salesforce and internal platforms. Built a daily ETL for ~150k records that cut refresh time ~70% and stabilized retries. Delivered bidirectional instance sync to keep data in parity across environments, with guardrails for schema drift and backfills. Added dashboards, alarms, and runbooks so ops could self-serve. Partnered with CRM and security to productionize access, secrets, and audit trails.

Shipped customer-facing microfrontend components that boosted feature engagement ~40%. Took on RCA for 10+ issues and closed fixes within the sprint, improving stability and cutting regressions with a focused test strategy.

Delivered data-intensive applications processing telecom tower signals and alerts. Reworked slow endpoints and background jobs, reducing p95 response times ~60% and lowering integration latency ~40% via event-driven flows. Tuned queries and indexes for measurable DB gains (~30%). Drove auth hardening and service boundaries to keep multi-tenant workloads predictable at scale.

Built data-heavy visualizations for millions of points and incorporated accessibility feedback, improving usability ~15% for key workflows.

Owned high-throughput APIs with a robust auth model and a test suite reaching ~95% coverage, cutting bug rates and keeping releases steady.