I specialize in Agentic AI, RAG Pipelines, and Production ML — the full arc from messy data to deployed, monitored systems. I've built pipelines that process 12,000+ documents daily, retrieval systems at 99.23% accuracy, and security tools that compress 40 hours of work into 30 minutes.
CS student at Multimedia University (graduating May 2026) — 3.74 CGPA, 4× Dean's List, Best Final Year Project.
Currently seeking an AI/ML Engineer role in FinTech, Healthcare, or AI-native companies.
| RAG Retrieval Accuracy | Docs Processed / Day | Red-Team Assessment | Intent Classification | Math Agent Boost |
🤖 Agentic Systems
Multi-agent architectures with LangGraph for adaptive reasoning, hybrid retrieval, and autonomous decision-making. Designed specialized agent teams (Query, Answering, Corrective) that collaborate dynamically.
🔍 RAG at Scale
End-to-end pipelines: document ingestion → semantic chunking → hybrid vector + BM25 search → knowledge graphs. Solved the "fragmented context" problem with multi-modal ETL and graph-powered retrieval.
⚙️ Production ML
Full lifecycle execution from data engineering to deployment and monitoring. ONNX-optimized inference, federated learning for privacy-preserving training, hyperparameter tuning at scale.
🛡️ AI Security & Red Teaming
Automated kill-chain pipelines with adaptive exploitation. Intelligence-driven vulnerability scanning with 50+ probes, human-in-the-loop approval gates for safe operation on production systems.
|
3-phase red-team orchestrator: Cartographer → Swarm → Snipers. Adaptive exploitation with Base64/ROT13/Unicode converters that learn from failed payloads. HITL gates before every phase. |
Event-driven system processing 12,000+ docs/day. Dual pipelines for PDF validation, repair, dedup, and structured extraction with fault tolerance. |
|
Full-stack RAG with VLM-powered PDF extraction, semantic chunking, and multi-agent hybrid retrieval solving the fragmented context problem. |
End-to-end AI document platform: LLMs + OCR + HITL review + real-time WebSocket updates. Full-stack Docker deployment. |
|
Designed the full system architecture. AI-powered payment fraud detection — 8 parallel rule indicators + LLM investigators for ambiguous cases. 4th place at Deriv Hackathon. 0.14s clean traffic, 12s deep investigation. |
Production RAG on AWS: CloudFront → API Gateway → VPC → EC2. Visual Knowledge Agent generating AI diagrams. SQS + Lambda async processing. |
|
95% accuracy, 0.90 F1-score on X-ray classification. Federated learning simulation for privacy-preserving model training. |
|
🏢 AI Engineer Intern — TM R&D (Telekom Malaysia), Aug 2025
Built a multi-stage, multi-modal pipeline using MinerU + Gemini 2.5 VLM for intelligent PDF-to-Markdown conversion. Designed a knowledge infrastructure with LightRAG achieving 99.23% retrieval accuracy. Boosted math agent accuracy by 18% via chain-of-thought and step-back prompting. Created evaluation providers integrated with Weights & Biases for automated model benchmarking. Compared Gemini, OpenAI, and Qwen-Plus across LLM tasks. Reduced token costs by filtering non-informative visual content during pre-processing.
🎓 BSc in Computer Science (Data Science) — Multimedia University, Malaysia
March 2023 → May 2026 · CGPA: 3.63 / 4.00 · 4× Dean's List · Best Final Year Project
Coursework: Statistics Calculus Discrete Math ML Algorithms Deep Learning AI OOP Database Management
I'm looking for an entry-level AI/ML Engineer role where I can ship production AI systems — particularly in FinTech, Healthcare, or AI-native companies. I care about building things that work at scale, not just in notebooks.
Open to roles involving: Agentic Systems · RAG Pipelines · ML Infrastructure · AI Security · Applied Research
If you're building something ambitious — let's talk.
Languages: English (Fluent) · Arabic (Native)