Building production-grade Agentic AI systems for retrieval, reasoning, and autonomous decision workflows.
What I've Built
Featured Projects
Production AI systems spanning autonomous agents, knowledge retrieval, aerospace simulation, and biometric security.
Hierarchical Multi-Agent Decision System for Wargaming
Production-grade C3 hierarchy multi-agent LLM system functioning as a Battle Management Assistant for real-time wargaming.
MARA - Multi-Agent AI Research Assistant
LangGraph-based multi-agent system that classifies intent, orchestrates parallel retrieval, validates citations, and generates structured research reports with human-in-the-loop review.
Knowledge Graph RAG for Aerospace Technical Documentation
Hybrid KG-RAG system answering multi-hop, relationship-dense queries about F-16 aircraft
Multi-Tenant Biometric Verification & AI Liveness System
Multi-stage AI liveness detection module with face quality check, passive depth analysis, active challenge-response and face embedding matching.
F-16 Technical Manual RAG Pipeline
Fully local, offline RAG system with layout-aware parsing, heading-aware chunking, two-stage retrieval (embedding + reranking), and evaluation module.
Behavioral Classification Pipeline for Clinical Interview Transcripts
Automated multi-label behavioral classification from clinical interview transcripts using generative AI and comparison with classical embedding + XGBoost approaches - evaluated against 26 behavioral labelled dataset.
Work History
Experience
Generative AI Engineer
April 2024 — Present · Islamabad, Pakistan
Key Contributions
- Architected hierarchical multi-agent wargame simulation with strategic planning, tactical agents orchestration, real-time spatial reasoning, and adaptive memory module.Live Demo
- Implemented & benchmarked hybrid retrieval architectures (RAG, LightRAG, Custom knowledge-graph) and evaluated on F-16 technical manual.
- Developed guardrail system for safe, grounded LLM responses in aviation domain
- Explored & benchmarked parsing models for technical manuals containing text, nested tables, and visual data.
- Built a RAG evaluation pipeline for comparing RAG, LightRAG, and custom knowledge-graph retrieval architectures on custom and synthetic dataset.
Past Independent Collaborations
Independent Engineering Work
MhTechFusion
Lead AI Developer
Cradle AI — Enterprise RAG Platform
Leading development of a Enterprise RAG platform with multi-tenant knowledge management.
Biometric Verification & AI Liveness System
Architected & built AI module: passive liveness detection, active challenge-response, and face embedding matching with optimization for deployment.
Skillion
AI Engineer
Behavioral Classification Project
Designed and implemented multi-label behavioral classification system for clinical interview transcripts using generative AI approaches, and compared with classical ML pipeline.
NCAI
Web Developer
Official Website Redesign & Development
Worked on the redesign and development of NCAI's official platform. Also engaged and participated in domain discussions regarding AI fundamentals.
Capabilities
Core Skills & Technologies
From agentic AI orchestration to production deployment — the full stack I use to ship AI systems.
AI Agents & Autonomous Systems
RAG & Knowledge Retrieval
LLM Engineering & Optimization
Backend & Infrastructure
LLMOps, Observability & Deployment
Front-End & Multimodal AI
Currently Exploring
Active Research Interests
Frontier AI topics I'm actively exploring, prototyping, and integrating into ongoing work.
Hybrid Graph-RAG Techniques
Exploring advanced hybrid retrieval architectures that combine structured knowledge graph traversal with dense vector search for improved multi-hop reasoning and context grounding in domain-specific corpora.
World Models for Autonomous Systems
Exploring world models for long-horizon planning and environment-aware autonomous agent systems.
Long-Horizon Agent Memory
Researching persistent memory architectures for agentic systems that maintain coherent context across extended task sequences without unbounded context windows.
Events & Publications
Conferences & Features
Baidu AI Wave Summit 2025
Overseas Developer RepresentativeSmart Heritage Application
Collaborated on an AI-powered cultural heritage guidance application using the ERNIE 4.5 multimodal vision-language model for 6 UNESCO World Heritage Sites in Pakistan.
- Built a multimodal "photo → recognition → multilingual narration" pipeline using ERNIE-4.5-VL and localized knowledge retrieval.
- Demonstrated real-world application of multimodal VLM, Ernie 4.5, for cultural heritage recognition.
Get in Touch
Let's Build Something
Available for Generative AI engineering, Agentic AI systems, and applied AI opportunities. Let's discuss!