
AI Engineer · Explainable AI Researcher
Imran Ashraf
Building AI you can question.
I build LLM and agent systems at Connectia Technologies and research explainable AI at VLCMatrixLab. Next stop: a fully funded PhD in Europe on AI that can explain its own decisions.
12 papers · Scientific Reports · IEEE Transactions on Consumer Electronics · Springer
About me
I'm an Associate AI Engineer at Connectia Technologies in Islamabad, where I work on LLM features, RAG pipelines and AI agents, along with the backend work that ships them: NestJS services, AWS Cognito sign-in flows and push notifications for a React Native app. Since 2023 I've also been a research assistant at VLCMatrixLab. I lead three projects there and mentor junior researchers. Most of my research asks one question: can a person check why a model made its call? My paper on explainable phishing website detection in Nature Scientific Reports, where I'm the corresponding author, is my clearest answer so far. I graduated in Software Engineering from The University of Lahore with a Gold Medal and a 3.80 CGPA. My final year project trained self-driving agents with multi-agent reinforcement learning, and I presented related work on vehicle platooning at the University of Wolverhampton, UK. Next, I want a fully funded PhD in Europe on explainable and agentic AI. If you supervise in that area, I'd like to hear from you.
Papers
12
First-author papers
4
Peer reviews
13+
CGPA · Gold Medal
3.80
What I work on
Explainable AI
Models whose decisions people can inspect and question.
LLMs, RAG & Agents
Retrieval pipelines and tool-using agents, built spec first.
Reinforcement Learning
Multi-agent RL (MAPPO) for autonomous driving and platooning.
Backend & Cloud
NestJS, TypeORM, AWS CDK and Cognito.
Selected Research
Experience
Featured Projects
Highlights & Recognition
Talks & Events
Education
Let's talk
I'm looking for a fully funded PhD in explainable or agentic AI in Europe, and I take on AI engineering work. If you supervise in this area or have a problem worth solving, I'd like to hear from you.