Please wait while we load the content
I'm Hanzala Irfan, an AI Systems Engineer with 4+ years building production pipelines, RAG workflows, and full-stack platforms that handle real data at scale.
I'm Hanzala Irfan, an AI Systems Engineer with 4+ years of experience building production AI pipelines and full-stack platforms that handle real data at scale.
Currently at AmroodLabs, I specialize in the Next.js/Node.js + AI stack, delivering document extraction pipelines, RAG workflows, and SaaS platforms serving thousands of users daily.
My focus is building systems that work under load — not just in demos. I've shipped AI pipelines achieving 95%+ accuracy, scaled platforms to 50K+ users, and maintained billing reliability at 99.9999% through comprehensive Stripe integration.
Building MARIA — an automated government dataset extraction system on AWS EC2 using Python OCR and Anthropic Claude, targeting 95%+ field accuracy across scanned PDFs. Also expanding SUCCESS.ai's AI infrastructure to support more advanced LLM integrations.
Technical Skills
AI-first. Full-stack capable. These are the tools I reach for when something actually needs to ship.
Two companies. Four years. The kind of work that either ships or gets cut.
Building Sportiv AI — an AI-powered RAG platform for sports contract compliance analysis. Processes contracts under one minute, checks against 2,759+ regulations and 1,700+ CAS/TAS decisions, and surfaces compliance risks across football, basketball, handball, volleyball, and rugby using ChatGPT and vector embeddings.
Architected asynchronous message queue infrastructure supporting distributed processing across AI pipelines and third-party integrations — ensuring reliable, scalable job execution under production load.
Architected MARIA — an end-to-end AWS EC2 pipeline that auto-downloads government datasets, unzips to S3, runs Python OCR, and routes structured data through Anthropic Claude for 95%+ accurate extraction.
Scaled SUCCESS.ai from 10K to 50K+ users — built every layer of the Stripe billing stack (subscriptions, usage-based, prorations, add-ons) with a 99.9999% reliability record.
Built BullMQ-orchestrated async job queues running sequential cronjobs across large document volumes — stable in production for 18+ months with zero data loss.
Integrated ChatGPT-4 and Anthropic Claude into production platforms — structured outputs, tool use, prompt versioning, and retry logic under real load.
Delivered internal API layers with Scalar docs and custom webhooks for third-party integrations (Apollo, analytics, payments) consumed by external clients.
Set up production monitoring and observability using Sentry for error tracking, Bugsink for self-hosted bug reporting, and Uptime Kuma for real-time service health — with CI/CD pipelines via GitHub Actions for automated deployments.
Wrote scraping pipelines in Nokogiri and Selenium extracting structured datasets at 98%+ accuracy — fed directly into downstream AI processing workflows.
Led a full Rails 3 → Rails 7 migration — updated gems, rewrote deprecated patterns, and kept the app functional throughout with no customer-facing downtime.
Integrated YouTube Analytics API to pull channel-level performance data and automated payment PDF downloads directly from YouTube Content Management System.
Built and maintained RESTful APIs in Ruby on Rails powering Hotwire and React frontends, including a Shopify-connected storefront.
Worked with Amazon Redshift for data warehousing — wrote queries and pipelines to surface analytics in the reporting dashboard.
Shipped responsive UI components across multiple client projects — collaborated directly with designers and QA on Figma handoffs and acceptance criteria.
EDUCATION
Where it all started.
Four-year program covering software engineering fundamentals, data structures, algorithms, and full-stack development. Graduated with a strong foundation in systems design and applied software architecture.
Final Year Project: DESIRETECK — a MERN stack web application where users configure and build their own custom computer systems through an interactive component selection interface.
Coursework in Software Design, Data Structures, Operating Systems, Database Systems, and Web Engineering.
Applied software engineering principles across team projects with real client requirements and delivery deadlines.
My Work
These aren't side projects. They're production systems handling real data for real users.
AI-powered RAG platform that lets athletes, agents, and coaches upload sports contracts and get instant analysis. Built a document ingestion pipeline with vector embeddings and ChatGPT to surface contract clauses, obligations, and risk flags — replacing hours of manual legal review.
Fully automated government dataset pipeline built on AWS EC2. Runs BullMQ cronjobs that download zip archives, extract to S3, run Python OCR on PDFs (both text-based and scanned), and route results through Anthropic Claude for structured field extraction — hitting 95%+ accuracy with no human intervention.
Scaled an AI-powered cold email SaaS from 10K to 50K+ users. Owned the full Stripe billing stack — subscriptions, usage-based plans, prorations, add-ons — achieving 99.9999% payment reliability. Integrated ChatGPT-4 for AI-assisted email generation and Apollo for lead enrichment across 700M+ B2B contacts.
YouTube Analytics & Payment Automation
Led a Rails 3 → Rails 7 migration on a live production app. Integrated the YouTube Analytics API to pull channel-level metrics and built automated payment PDF downloads from YouTube CMS — both replacing manual workflows that previously required engineering time each month.
Web Scraping & Data Extraction Pipeline
Built scraping pipelines in Nokogiri and Selenium that extracted structured datasets at 98%+ accuracy from sites with complex DOM structures and JavaScript rendering. Data was cleaned, validated, and stored directly into PostgreSQL for downstream processing.
Refinery CMS Optimization
Maintained and stabilized a Refinery CMS deployment for TiderlandEMC — triaged 50+ production bugs, optimized slow queries, and worked with the content team to streamline their publishing workflow. Hosted and deployed on AWS.
Every project above ran — or still runs — in production. Real users, real data, real failure modes. That's the only environment where engineering decisions actually get tested.
Get In Touch
If you're building something that needs real AI engineering — or need someone who can own a full stack — I'm available.
I check messages daily. If it's a serious inquiry, I'll respond within a few hours.