class AIEngineer:
name = "Muhammad Abdullah"
location = "Haripur, Pakistan"
role = "Junior AI Engineer @ Revnix"
education = "B.Sc. Artificial Intelligence — PAFIAST (Final Year)"
focus = [
"Agentic Workflows & Multi-Agent Orchestration",
"MLOps — DVC, MLflow, reproducible pipelines",
"Computer Vision — YOLOv11, TFLite, edge deployment",
"Low-Latency Voice AI — WebRTC, LiveKit, ElevenLabs",
]
currently_learning = [
"Advanced LangGraph state machines",
"Distributed inference architectures",
"Edge AI optimization for mobile deployment",
]
contact = "cabdullahbokan@gmail.com"|
Road defect detection system for real-time infrastructure diagnostics. Trained YOLOv11 on a custom-labeled dataset achieving ~94% mAP. Exported to TFLite for on-device inference in <40ms. Automated PDF reporting pipeline cuts manual inspection time significantly. Stack: Python · YOLOv11 · TFLite · OpenCV · FastAPI View Source Code → |
Stateful multi-agent systems with cyclic reasoning and autonomous task execution. Built graph-based workflows with conditional routing and memory persistence. Supports parallel branch execution for latency-sensitive pipelines. Stack: LangGraph · LangChain · Python · OpenAI API · FastAPI View Source Code → |
|
Real-time conversational voice agents integrated with telephony and WhatsApp. End-to-end pipeline: STT → LLM reasoning → TTS in under 800ms round-trip. Integrated LiveKit for WebRTC transport and ElevenLabs for natural synthesis. Stack: Python · LiveKit · ElevenLabs · WebRTC · FastAPI · Twilio View Source Code → |
Reproducible, production-grade ML pipelines with full experiment tracking. Versioned datasets and model artifacts with DVC. Tracked 50+ experiments via MLflow for clean model comparison and rollback. Containerized with Docker for consistent cross-env execution. Stack: Python · DVC · MLflow · Docker · Scikit-Learn · PostgreSQL View Source Code → |
