I build AI systems that run unattended.
Autonomous agents that watch, decide, produce and publish — while nobody is at the keyboard. Content engines posting to four platforms a day, CRM agents answering hundreds of customers a week, infrastructure that heals itself when it breaks at 3am.
Case files
Each of these runs in production today. Built end-to-end: architecture, agents, pipelines, deployment, monitoring.
ReelMachine
liveagentic pipelinePaste a wildlife clip link — an AI honey-badger presenter watches the footage, writes its own commentary, speaks it, lip-syncs on GPU, edits the reel and posts it to YouTube, Instagram, Facebook and TikTok. One tap. Vision models pick the framing, the highlight moment and where the presenter stands; a measured audio chain keeps the voice on top; QC gates every render before publish.
BRINK
liveautonomous networkA news-shorts network that runs itself: scouts breaking stories, sources real footage, writes scripts, renders branded vertical video and publishes to YouTube, Instagram, Facebook and TikTok across three languages — English, Hindi, Spanish — on a fixed daily schedule, unattended. A quality gate inspects every video (branding, black frames, audio) before it ships.
STEMEX CRM automation
productionemploymentAt STEMEX (Norwich), designed and shipped the AI layer of the business: an autonomous customer-care agent handling 200–250 inbound emails and SMS a week at an ~80% automation rate, and a marketing agent that generates creatives and feeds ad campaigns. Rebuilt the CRM end to end, consolidating ~25 fragmented pipelines into clean customer journeys, and migrated 34,000+ transactions off legacy EPOS into a single analytics view.
Movie with Eman
livefull-stackA private streaming platform for watching films together: multi-gigabyte uploads, transcoding, synchronized playback over websockets, invite links, and a tuned nginx/Node delivery path that keeps slow residential connections from stalling large streams.
Control Room
liveinfrastructureThe operations layer under everything above: a fleet dashboard plus self-healing watchdogs — services that restart themselves, channels that reconnect, disk/token/queue monitors that alert only on state change. Designed so the answer to "is it still running?" is always yes.
Where I've worked
AI & Automation Engineer · STEMEX
- Own the company's agentic-AI stack end to end — customer care, marketing, monitoring, reporting
- Replaced a dedicated support function with an autonomous agent handling 200–250 contacts/week at ~80% automation
- Rebuilt the CRM end to end: ~25 pipelines consolidated, 1,300+ dead records cleaned; re-engagement drove +20% bookings
- Migrated 34,000+ transactions off legacy EPOS; a verified 91% attendance rate feeds footfall and revenue analysis
Freelance Data Scientist · Fiverr
- 20+ delivered ML/analytics projects across healthcare, agriculture, retail, tech — 4.9/5 rating
- Full pipelines: extraction, ETL, feature engineering, modelling, deployment, BI dashboards
Computer Vision Intern · KICS, UET Lahore
- YOLO/OpenCV detection pipelines; medical-imaging classification with ResNet/VGG transfer learning
Under the hood
Agents & LLM engineering
Claude Code agent development, reusable Skills, tool-use / function calling, multi-agent orchestration, prompt engineering, RAG
Automation
Python, n8n, Zapier, API integration, CRM (GoHighLevel), webhook & pipeline design
ML & computer vision
PyTorch, TensorFlow, YOLO, OpenCV, NLP — MSc Intelligent Vision (Lincoln, 2023), Grad-CAM explainability research
Cloud & MLOps
AWS (SageMaker, S3, Lambda) — Certified Cloud Practitioner · Docker, MLflow, Git, CI/CD
Media engineering
ffmpeg filtergraphs, GPU render pipelines, chroma-key compositing, loudness-normalised audio chains
Data & BI
SQL, Pandas, statistical analysis, Power BI (Microsoft Data Analyst Specialization)
Building something that should run itself?
Open to AI engineering and automation roles — remote or UK.