I design and build agentic AI systems, real-time voice agents, and enterprise automation platforms. My focus is reliable AI workflows that reason, use tools, manage memory, and take action in complex environments.
Available for senior AI engineering work
A testing platform for voice AI. Synthetic personas call a bot through scripted scenarios, every run is scored against evaluation criteria, and results roll up into session and organisation reports — so a prompt or model change gets caught before customers hear it. Runs scheduled across timezones with exportable reports.
ReactTypeScriptSupabaseAWS AmplifyLLM evaluation
A browser-based calling app where two people speak different languages and hear each other translated inside the same live call. Twilio call control, streaming speech-to-text, translation and neural TTS stitched into one near-real-time loop, with live transcripts, consent-aware recording, wallet billing and per-country rating behind it.
Next.js 14FastAPITwilio VoiceSoniox STTGoogle Translate + TTSFirebaseCloud Run
Real-time conversational agents on customer-facing phone and web channels: barge-in handling, tool calls into business systems, warm transfer to humans, and the latency budget work that keeps a conversation feeling like a conversation.
VapiElevenLabsPipecatTwilioAsterisk
An OCR pipeline that reads invoices and pulls structured fields out of them, replacing manual data entry in a high-volume document workflow. Trained and served on SageMaker, containerised for repeatable deployment.
PythonAWS SageMakerOCRDocker
Facial emotion recognition adapted from Western benchmark datasets to Pakistani dramas and talk shows — a study in what happens to a vision model when the domain shifts underneath it, and how to close the gap without collecting a dataset from scratch.
PyTorchComputer VisionDomain Adaptation
I’ve built and scaled AI systems for telecommunications, healthcare, customer support, and enterprise automation — from conversational AI and IVR to multi-step orchestration and low-latency voice infrastructure supporting thousands of concurrent interactions.
I also build AI testing harnesses: validating conversational flows, tool calls, and multilingual interactions through regression testing, load testing, and end-to-end evaluation. The goal is fewer hallucinations, lower latency, and systems that hold up in production.
Founding engineer at CloudCX, helping grow the business from £0 to £1 million in annual recurring revenue (ARR). Built agentic AI systems, real-time voice agents, and enterprise automation.
Helped scale Metric from 0 to 100,000 app downloads as Head of Engineering. Joined as the second employee in the new tech team and helped grow engineering from 1 to 10 people. Established hiring and onboarding processes, improved releases and CI/CD with Jenkins, and resolved Firestore database performance issues.
DeepLearning.AI · Jul 2021
Credential ID: AZTENR9E8JFU
Udacity · Sep 2020
Credential ID: YX3PJDKM
Udacity · Aug 2020
Coursera · May 2020
OpenDataScience [ods.ai] · Sep 2019
Coursera · Aug 2018
Away from work, I enjoy chess, padel, table tennis, and badminton. I’m a big fan of productivity books and podcasts, and I love travelling and exploring the world. These interests have been part of my life for a long time.
That curiosity carries into everyday life, too. I enjoy exploring new AI tools like OpenClaw, Instinct, and Muse by Meta AI, and finding ways to automate my daily routines with AI.
I’d love to share what I’ve learned with junior engineers, help them build their skills, and support them as they grow in their careers. If you’re starting out and looking for guidance, get in touch.
Open to senior AI engineering work — voice agents, LLM systems, and the infrastructure around them.