Shipping AI products, modernizing old systems, and building software that holds up.

A production AI SaaS that converts PDF bank statements into Excel, CSV, and QuickBooks-compatible exports for accountants, bookkeepers, lenders, and finance teams.

• Built a Cloudflare-native full-stack architecture with a web Worker, queue consumer, D1 database, R2 file storage, KV progress state, and Stripe billing.
• Designed an AI document pipeline with PDF extraction, structural fingerprinting, cached parser specs, LLM-generated structured output, deterministic interpretation, and validation gates.
• Added reliability work around balance reconciliation, golden datasets, field-level evaluation plans, review queues, model testing, and cost controls.
• Built monetization and operating systems: subscriptions, pay-per-use, multi-currency pricing, webhooks, transactional email, product analytics, SEO snapshots, and India-market experiments.
• Developed programmatic SEO surfaces including bank-specific landing pages, conversion pages, calculators, resource pages, JSON-LD, sitemap generation, and LLM-facing documentation.

Remote engineering work for a UK-based education services company, spanning old PHP systems, WordPress migration, order flows, portals, email templates, and infrastructure.

• Developed frontend work for a new order system and contributed across order form, client portal, researcher portal, and shared vendor libraries.
• Helped migrate a legacy CMS with direct MySQL access into WordPress-backed content, custom themes, plugins, and API wrappers.
• Improved WordPress search performance by 200% through custom search and migration work.
• Raised mobile PageSpeed from 60 to 100 through performance optimization on production-facing pages.
• Shipped practical fixes across production workflows: staff notifications, message flows, email templates, report downloads, access control, responsive behavior, and deployment configuration.

A time-sensitive consumer alert product across iOS and WeChat Mini Program, built for users who needed timely city-specific notifications.

• Built city-based alerting and subscription workflows across 23 Chinese cities.
• Shipped Apple auto-renewable subscription monetization with five approved plans from one month to one year.
• The iOS app alone reached approximately 57.5K App Store downloads, 13.5K App Store subscription units, $23.9K gross subscription sales, and $20.1K developer proceeds.
• Iterated from v1.0 to v1.3.2 in roughly 11 weeks while expanding supported cities and improving stability.
• Grew through ASO, social media, referral campaigns, and time-sensitive product-market fit in a scarce appointment market.

A portfolio of focused iOS utilities published under Pure Insight Technology Ltd while BankScanPro is being built as the main SaaS product.

• Published and maintains public support, privacy, terms, and product pages for a catalog of 14 apps.
• Apps cover practical niches including recipe pricing, contractor estimates, vehicle records, homeschool logs, hair color formulas, health tracking, personal bookkeeping, and AI garden redesign.
• Use as evidence of shipping cadence and mobile product breadth, not as traction proof until stronger public metrics exist.

Earlier work that built the range behind the current product engineering practice: frontend scale, hardware-adjacent performance, and security research.

• At Intel, worked on Quick Assist Technology for cryptography and compression, including QAT NGINX performance testing and VDO benchmarking automation.
• At Tencent MIG, built frontend pages and components for the Yingyongbao open platform and related developer surfaces.
• In a university security research group, researched deep learning applied to WAF request classification and built Python security tooling.