AI Engineer
I build production AI systems — LLM pipelines, distributed backends, real deployments on GCP. Not notebooks. Not demos. Things that run under load and stay up.
Selected Work
Production systems with real users and real load — not portfolio demos.
AI image & video generation platform for iOS. Users generate personalised photos with preset backgrounds, create video ads, try on clothes virtually, and share via a custom WhatsApp keyboard — all powered by per-feature AI pipelines I built from scratch.
Designed a distributed job system with retry logic, idempotency, and locking so concurrent AI generation tasks never lose state. Handled full GCP deployment with IaC and automated CI/CD from day one.
LLM-powered outreach system deployed in production. Reduced manual email effort by 90%. Handles sequencing logic, personalisation, and delivery — all driven by LLM APIs.
Built with full idempotency, retry logic, and automated CI/CD on GCP Cloud Run — handles failures gracefully and never sends duplicate emails under load.
Real-time physiotherapy app using MediaPipe pose estimation. Tracks reps, detects form errors in live video, syncs session data between physiotherapists and patients.
View on GitHub ↗Full RAG pipeline querying 1,000+ pages with accurate answers in under 2 seconds. FAISS vector retrieval with a React frontend — reduced manual search time by 70%.
Collected and processed 1,000+ records via scraping and preprocessing. Built Regression and SVM models to forecast seasonal demand and analyse pricing trends.
View on GitHub ↗Urdu sarcasm detection, English-Urdu machine translation using RNN/LSTM, and Phi-3 fine-tuning. End-to-end pipeline from preprocessing through classification and evaluation.
View on GitHub ↗K-Means clustering to group 2,500+ students by domain and batch. Generates automated PDF seating plans and faculty assignments for university exams — zero manual allocation.
View on GitHub ↗Binary image classification using HOG, LBP, GLCM and VGG19 deep features. PCA dimensionality reduction with Linear SVM, evaluated via 10-fold cross-validation.
View on GitHub ↗Education
Experience
Capabilities
Contact
Open to remote freelance projects and full-time roles at ambitious companies. If you need AI systems that work in production — not just in a notebook — let's talk.