Kitco — Frontend Engineering, Financial Data UX & Cloud Reliability

Overview

Kitco is a high-traffic financial and precious metals platform combining real-time pricing, market charts, editorial content, media, and community features.

I worked across multiple services with a focus on frontend delivery, financial data experiences, performance, production reliability, and cloud scalability.

Scope of engagement

Frontend engineering, financial data UX, API performance, cache strategy, and platform reliability across core web properties and supporting services.

What I owned

• Delivered frontend work on a large Next.js application serving investor-facing pricing and market-data experiences.
• Built chart-driven financial interfaces designed to surface live market information clearly and reliably.
• Improved caching and request-handling strategy to reduce backend pressure on high-traffic market-data paths.
• Built symbol import, indexing, and synchronization workflows supporting charting and market discovery experiences.
• Diagnosed and resolved production issues affecting availability, performance, and runtime reliability under live traffic.
• Improved Kubernetes-based service reliability through readiness, scaling, rollout-safety, and traffic-handling enhancements.
• Onboarded engineers into the frontend architecture, deployment workflow, and production model.

Financial data and charting

I built and operated investor-facing market-data experiences: pricing pages, chart integrations, symbol-driven workflows, and supporting services responsible for keeping financial data reliable and usable in production.

This included chart-heavy interfaces, frequently updated pricing experiences, symbol ingestion and synchronization, and reliability improvements in the systems behind searchable market data and chart integrations.

Performance and reliability

Rewrote indexed CMS queries and fixed cache invalidation so Cloud SQL CPU fell from 100% to about 20% across two market opens; client-reported query time fell from about 1.5 s to 0.5 s.

Worked directly on production issues across the frontend/platform boundary, improving cache behavior, service readiness, rollout safety, and runtime stability.

Selected engineering wins

• Upgraded the search platform from end-of-life Solr 8 to Solr 9 with zero downtime: search configuration baked into a reproducible container image, the full cutover rehearsed in dev first, and production monitored live for hours before handoff.
• Root-caused a production outage where the cache layer served stale HTML error pages for JavaScript assets after deploys: fixed the cache policy and the deploy-time purge to prevent caching these asset errors, and posted the structured postmortem within minutes of the incident ticket, with the fix already in place.
• Fixed an accumulating fleet of about 50 saturated Next.js pods by adding HTTP readiness checks: stuck pods timed out at 3 s while healthy pods answered in 2–6 ms.
• Normalized caching across Next.js and Varnish with a 30 s API TTL and 1 h stale-while-revalidate grace, serving stale responses when the backend failed.
• Authored and completed an 11-task cloud cost-reduction program for a bill trending toward about $14.5k/month: removed 30-second polling of once-daily data, right-sized a 17-node cluster, and tuned autoscaling and resource limits.
• Operated the workloads on a regional three-zone GKE cluster with separate node pools by workload; the general-purpose pool was configured to autoscale from 5 to 60 nodes in the July 2026 estate audit.

Platform scale

The Google Analytics snapshot I recorded on March 27, 2026 showed 24M active users in its report and 17,758 active users in the last 30 minutes.

Outcome

The work delivered a zero-downtime Solr 9.10 cutover, resolved cached-asset outages and forum 503 bursts, cut forum pod boot from 5–6 minutes to about 60 seconds, and reduced Cloud SQL CPU from 100% to about 20%.