Software Engineer · AWS · Dallas, TX
Hey, I'm Moeez.
A software engineer at AWS and a first-generation kid from a small village in Pakistan. These days I build AI agents that read decades-old code and help carry it into the present — teaching machines to understand the systems the world quietly runs on.
currentlybuilding modernization agents at AWS
What I do all day
I teach machines to read code older than I am.
Not a diff tool — an agent. It reads a mainframe program, plans its own approach, calls tools to recover the rules hidden inside, rewrites it for today, then a second agent reviews the result. Here's one working, live.
IDENTIFICATION DIVISION.PROGRAM-ID. INT-CALC.WORKING-STORAGE SECTION.01 WS-BALANCE PIC 9(7)V99 COMP-3.01 WS-RATE PIC 9V9(4) COMP-3.01 WS-INTEREST PIC 9(7)V99 COMP-3.PROCEDURE DIVISION. IF WS-BALANCE < 0 MOVE ZERO TO WS-INTEREST ELSE COMPUTE WS-INTEREST = WS-BALANCE * WS-RATE END-IF.
BigDecimal interest(BigDecimal balance, BigDecimal rate) { // recovered rule: no interest on an overdrawn account if (balance.signum() < 0) return BigDecimal.ZERO; return balance.multiply(rate) .setScale(2, RoundingMode.HALF_UP);}
- Parse COBOL structureread_source()
- Detect COMP-3 packed decimalsanalyze_types()
- Recover business rulesextract_logic()
- Generate modern Javasynthesize()
- Verify behaviorrun_tests()
How I got here
Four homes, one throughline.
I was born in a small village in Pakistan.
- Born in a small village
- First-generation learner
- Where the whole thing starts
Then I grew up in Amsterdam.
- Grew up here
- A long way from a small village
- Two worlds, one kid
I studied in Indiana.
- DePauw — Physics → Computer Science
- 6 software engineering internships
- Entrepreneurship Group President
- Resident Assistant · Student Senator
And now I build in Dallas.
- Software Engineer, AWS Transform
- Agentic AI for modernization
- Speaking at 2 conferences in 2026
Talks & Speaking
Presenting the work where the field gathers.
No recordings up yet — talks move here once delivered. First one: Sep 3, 2026.
Click a talk to expand.
Writing
Thinking out loud about agentic AI and legacy systems.
- 01An AI Agent Already Escaped Kubernetes. Is Your Cluster Next?In May 2026, an LLM harness with no human in the loop broke out of a container, replayed a stolen service-account token, and dumped a Kubernetes cluster's entire Secret store. Kubernetes was never built to contain that. Here's the blast radius, and how to close it.
- 02The 31% Problem: How to Actually Use Claude Code WellFaros AI tracked 22,000 developers through 2026 and found pull requests merged with no review at all are up 31.3% as AI coding tools spread. The model isn't the problem, every tool ships the same one. It's a handful of structures most people never turn on. A comprehensive, visual walkthrough of all of them.
- 03The Three Ways AI Agents Fail (and How to Fix Yours)MIT found 95% of enterprise GenAI initiatives deliver no measurable return, and it's almost never the model's fault. Tool connectivity, memory, and multi-agent coordination are where systems actually break. Adapted from my talk at Commit Your Code 2026.
- 04COMP-3 and the Quiet Danger of Modernizing MainframesThe hardest part of moving a COBOL program to a modern language isn't syntax — it's the arithmetic. A single misunderstood packed-decimal field can silently change a customer's balance. Here's why, and how agents can get it right.
- 05The 45% Problem: Catching Vulnerabilities in AI-Written CodeTwo years of smarter models haven't moved the rate — roughly 45% of AI-generated code still ships a security flaw. The dangerous part is that it looks perfect. Here's the setup I use to catch it before it merges.
- 06The Agent-Ready Audit: Redesigning Your Week Around DelegationThe mechanical 80% of your week no longer needs to be your burden. A one-page audit for finding the first three tasks to hand to an agent — and the three-question test for what to keep.
Each piece is cross-posted and citation-ready — reshared to LinkedIn, Medium, and Substack.
Open Source
Building in the open — where the work is verifiable.
The most durable evidence isn't a claim, it's a commit. I'm consolidating my work into one flagship, continuously-evolving project — tooling that makes legacy modernization tractable for anyone, not just the team that owns the mainframe.
The modernization toolkit
An open framework for agent-driven modernization: read legacy code, recover the business rules inside it, and translate it forward with behavior locked by tests — plus the evaluation and observability layer to prove the migration is safe.Name and first release coming soon.
- In development
- First release
- Deployed
- 0
- stars earned
- 0
- forks
Recognition
Features, honors, mentorship, and programs — with the receipts.
Speaking2 accepted conference talks in 2026 — see the Talks section ↑- 2026Mentor100% rated across 9 ADPList reviews Communication, motivation, problem solving, subject knowledge — every category · See reviews ↗
- 2026Mentor5/5 across 8 Topmate testimonials 16 sessions on resumes, interviews, and career strategy · See testimonials ↗
- 2025–26Mentor4 mentee shoutouts on LinkedIn — Amazon, AWS, and McKinseyPost 1 ↗ · Post 2 ↗ · Post 3 ↗ · Post 4 · mainframe modernization ↗
- 2024
- 2024
- 2023
- 2023
Growing — media, external reviewing, and judging to be added as they land.