Aleppo Governorate, Syria
Senior AI & Backend Engineer with 4+ years building production-grade, event-driven systems and applied AI at scale. I don't just train models — I architect and provision the observable infrastructure that runs them reliably in production, from immutable Kubernetes clusters to Kafka-driven pipelines handling high-throughput workloads. Most recently, I engineered an end-to-end Django + Claude Vision VLM pipeline for an enterprise client that collapses a multi-stage OCR/LLM process into a single structured-JSON call, with confidence-scored routing sending only uncertain fields to human reviewers. I've also designed and shipped three live, self-hosted AI products end-to-end: → NeuroSeek AI — an agentic research platform with hybrid RRF-fused RAG (pgvector + keyword search) that plans multi-step searches, audits its own answer coverage, and streams fully cited reports. → PulmoSight — a doctor-in-the-loop clinical AI that classifies 18 chest pathologies with Grad-CAM heatmaps, gated by a safety layer that keeps every AI finding under human sign-off. → ROTECHAI — a multilingual AI studio platform right-sized to run without a GPU or paid LLM API. I'm a polyglot backend engineer — expert in Django and Go, with FastAPI powering dedicated AI-serving microservices — and I build for scale and reliability: Terraform and Ansible for infrastructure-as-code, Apache Kafka & RabbitMQ for messaging, Kubernetes for orchestration, and Prometheus/Grafana for observability. Selected impact: - Cut PostgreSQL query time from 2,000ms to under 50ms (40x) via advanced indexing & query analysis - Improved API performance 60–70% with Redis caching; reduced database load by 65% - Deployed a computer-vision model (AUC-ROC 0.997) as a live production API for medical imaging - Built an HR attrition model (87% accuracy) estimating $225K–$450K in annual retention savings, plus forecasting pipelines on 1M+ transactions EU Blue Card eligible and open to relocation across Germany / Netherlands, or fully remote roles within EU time zones. Let's build something real.
Designing and deploying production-grade AI systems — from agentic RAG to NLP and computer vision — while architecting the event-driven, observable infrastructure that runs them in production. - Engineered an end-to-end Django + Claude Vision VLM pipeline for an enterprise client (Austria), transforming scanned Arabic official documents into formatted DOCX output — collapsing a multi-stage OCR/LLM stack into a single VLM call returning strict JSON, and directly manipulating raw OpenXML table structures to preserve exact fonts and layouts. - Designed a confidence-scored human-in-the-loop routing workflow: clean VLM outputs pass instantly, while uncertain fields (<0.85 confidence) route to role-based reviewers with full audit logging. - Provisioned secure, multi-node infrastructure as code with Terraform and Ansible, deploying decoupled FastAPI inference microservices across Kafka/RabbitMQ pipelines, with full observability via Prometheus/Grafana and pgBackRest for point-in-time recovery. - Built a chest X-ray classifier (EfficientNetV2S) achieving AUC-ROC 0.997 / F1 0.95 via Django REST API, a sales forecasting pipeline (Facebook Prophet) on 1M+ transactions, and an HR attrition model reaching 87% accuracy with $225K–$450K in estimated annual retention savings.
Designed and delivered scalable, polyglot backend systems and event-driven architectures for production applications. - Built and shipped secure REST APIs — via Django REST Framework and Go — with JWT/OAuth2 authentication, versioning, and OpenAPI documentation, plus GraphQL endpoints for flexible client queries. - Reduced PostgreSQL query response times from 2,000ms to under 50ms (40x), cut database load by 65%, and improved API latency 60–70% through strategic Redis caching. - Designed real-time systems (WebSockets, Django Channels, Redis) for live ride-matching and location streaming. - Integrated Stripe, PayPal, and Razorpay with secure webhook handling and transaction processing. - Deployed Go microservices for latency-critical, high-concurrency components (WebSocket hubs, event consumers), with RabbitMQ/Kafka decoupling asynchronous processing and pgBackRest/pgBadger strengthening PostgreSQL backup and query analysis. - Deployed backend services using Docker, Kubernetes, NGINX, and AWS infrastructure for high availability and scalability.
Delivered cross-platform Flutter applications across Android, iOS, and Desktop with clean architecture, advanced state management, and deep backend integration. - Applied BLoC, GetX, and Riverpod state management with modular data/domain/presentation architecture across all projects. - Integrated the full Firebase suite: Authentication, Firestore, Storage, FCM push notifications, and ML Kit. - Built real-time Flutter clients over WebSockets (Django Channels) with live map tracking, polyline routing via Google Maps SDK, and offline-first persistence using SQLite and Hive.