Junaid Hassan
Independent AI Researcher & Founder, Codoplex
I build production AI systems — large language models, retrieval-augmented generation, and agentic architectures — and research how to make them more reliable. Published in Neural Processing Letters. Currently pursuing doctoral research opportunities in AI/ML.
About
I hold a BSc in Telecommunications Engineering (UET Taxila) and an MSc in Computer Science with a specialization in AI and Deep Learning (University of Gujrat). My MSc thesis proposed a gated recurrent neural network architecture for text classification, later published in Neural Processing Letters (Hassan & Shoaib, 2020) after peer review.
Over a decade of software engineering, I’ve moved from full-stack web development into building production systems around large language models — retrieval-augmented generation, multi-agent orchestration, fine-tuning, and multi-backend inference infrastructure — as founder of Codoplex. I’m now applying that practitioner’s perspective to doctoral-level research in AI/ML.
Research Focus
🧠
Agentic AI Systems
Multi-agent orchestration, tool-use, and inter-agent communication for complex, multi-step tasks.
🔎
Retrieval-Augmented Generation
Grounding LLM outputs in retrieved context to reduce hallucination and improve reliability.
⚙️
Efficient Adaptation
Parameter-efficient fine-tuning (LoRA/PEFT) and multi-backend inference routing for cost and reliability.
Publications
Hassan, J., & Shoaib, U. (2020). Multi-class Review Rating Classification using Deep Recurrent Neural Network. Neural Processing Letters, 51, 1031–1048.
Technical Writings
From Zero to Agents — self-published two-volume technical course (Leanpub, 2026). A from-first-principles AI/ML curriculum covering foundational mathematics through transformers, fine-tuning, and agentic systems, with every implementation independently derived and empirically verified.
Let’s connect
Message me on WhatsApp or send an email.