Manal H.

• Architecture of a real-time middleware analyzing and auditing responses from multiple LLM providers
• Design of hallucination detection and risk content flagging mechanisms
• Complete frontend React development: real-time dashboards, audit interfaces, and multi-tenant client portals
• Design of a multi-client integration layer for B2B SaaS distribution
• Management of the complete product lifecycle, from design to production and iterative evolution
• Development of REST APIs (FastAPI) and streaming pipelines for low-latency response interception

• Real-time middleware handling 100K+ requests/day with p95 latency < 250ms
• 91% precision and 87% recall in multi-LLM provider hallucination detection on internal benchmarks
• Multi-tenant architecture supporting isolated client workspaces without inter-tenant data leakage
• React dashboard offering live audit visualization with refresh latency < 500ms
• End-to-end observability (OpenTelemetry) and CI/CD pipelines enabling rapid and uninterrupted production iterations

Context:

Main Missions:

• Architecture of AI microservices with REST APIs (FastAPI) for model ingestion and exposure
• Development of end-to-end ML pipelines on GCP (Vertex AI, BigQuery, Cloud Run)
• Design and deployment of AI agents and GenAI solutions for workflow automation
• MLOps Pipelines: automated retraining, monitoring, drift detection
• Kubernetes orchestration (GKE) and Docker containerization

Measurable Results:

Context:

Achievements:

• Architecture and deployment of scalable ML infrastructures on GCP
• Integration of backend APIs for prediction exposure
• Computer Vision Models: object detection, image classification

Results:

• Optimization of ML pipelines with a 40% performance improvement
• 15% accuracy gain on object detection models