AI Engineer who ships GenAI solutions with enterprise clients. IEEE-published researcher. I build with AI-native tools and believe the best engineers will be the ones who learn to work with AI, not around it.
Currently: Software Engineer @ Turing — building agentic automations and backend services on GCP.
Previously: Fine-tuned LLMs for Apple, Meta, and Bytedance. Delivered GenAI integrations for Fortune 500 clients at Royal Cyber. Published at IEEE EMBC 2025 on explainable AI for medical imaging.
AI/ML Agentic Systems · RAG · RLHF · LLM Fine-Tuning · PyTorch · LangChain
Backend Python · FastAPI · PostgreSQL · Redis · Microservices
Cloud GCP · AWS · Docker · CI/CD
Enterprise Mulesoft · Salesforce · Apache Kafka · IBM ACE/MQ
AI Tools Cursor · OpenAI Codex · Claude Code · GitHub Copilot
Production-ready starter repo with CI/CD workflows, test coverage enforcement, incident runbooks, pre-commit hooks, and AI coding tool configs (Cursor, Claude Code, Codex, Copilot). Clone it, customize it, ship.
What's inside
- 3 GitHub Actions workflows — PR checks with Ruff auto-fix, test coverage enforcement (posts PR comments with exact test guidance), merge readiness gates
- Pre-commit hooks — Ruff, syntax checks, YAML/JSON/TOML validation, private key detection, branch protection
- Issue templates — Bug reports, incident reports, infrastructure issues
- Incident runbook — Severity classification, triage checklists, diagnostic commands
- AI tool instructions — AGENTS.md, CLAUDE.md, .cursor/rules/, .claude/rules/, .github/copilot-instructions.md
A local FastAPI tool for disciplined monthly investing on the Pakistan Stock Exchange. Enter an amount and it pulls live PSX prices, then computes exactly how many whole shares of each holding to buy to hit your target weights.
What's inside
- Live price proxy — async
httpxfetch from the PSX data portal with an intraday -> end-of-day -> cached fallback chain and staleness flags - Buy-plan math — whole-share flooring with an optional leftover-cash optimizer that nudges the book toward target weights
- Purchase history & positions — record actual fills, then track average cost, market value, and unrealized P/L at live prices
- No-build frontend — one self-contained HTML page (Tailwind CDN + vanilla JS) served by the Python backend
- Privacy by design — personal config and history stay in a git-ignored private repo; only the app and a generic example ship publicly
- Tested & CI'd — pytest unit tests, Ruff, and GitHub Actions
A source-backed catalogue of AI-writing markers with a zero-dependency Python checker. Built as a transparency and self-editing aid — explicitly not an AI detector, since detection is probabilistic and biased against non-native and academic writing.
What's inside
markers.json— 6 categories, 87+ markers (overused vocab, formulaic transitions, cliché phrases, structural/rhetorical tells, punctuation/format, statistical), each with era tags and source idscheck.py— zero-dependency CLI that scans any file/stdin, reporting marker hits by category with per-1000-word density- Real burstiness metric — stdev/mean of sentence length; honest about not faking a perplexity score
- Sourced — distilled from Wikipedia's WikiProject AI Cleanup guide, 2026 detector research, and the peer-reviewed Patterns false-positive study
- Open data — MIT (code) / CC0 (dataset)
- Shipped GenAI solutions for Apple, Meta, Bytedance, and Fortune 500 clients
- Built AI agents that reduced client workloads by 60%
- Led a distributed team of 10+ engineers across 3 continents
- IEEE EMBC 2025 — OptiGuard: Explainable Glaucoma Classification
- IBM Certified Cloud Automation Sales Consultant
- Mulesoft Certified Developer — Level 1
- Microsoft Certified: Power Platform Fundamentals (PL-900)
Open to GTM Engineering, Forward Deployed, Solutions Engineering, AI, and PhD opportunities.