I'm Asaad Cheema, Applied AI Engineer at Mowi, embedded with the S&M Digital Solutions team. AWS-certified Data Engineer and Solutions Architect, with PhD-level depth in federated learning, computer vision, and LLM-driven workflows. I don't only design solutions; I deliver them, from research notebook to production cloud.
AC
Currently at Mowi · S&M Digital Solutions
Working with a sharp Mowi team to design a private internal GPT, augmented with task-specific agents (retrieval, forecasting, customer-context, decision-support) orchestrated to lift sales productivity.
Building autonomous agents that read, classify, and reason over complex commercial contracts. They surface risk, obligations, and renewal triggers so legal and commercial teams act on insight, not paperwork.
Modernizing data flows from production through commercial, unifying signals into a single decision surface so domain experts can act on AI-informed insight in minutes, not weeks.
Embedding ML and LLM-driven copilots into existing operational workflows, so teams stop hunting through dashboards and start receiving the insight where the work happens.
I spent four years at NTNU earning a PhD in Topology-Aware Machine Learning for IoT, with a research stint at Imperial College London. In parallel, I shipped production ML at Wrist-Shot (luxury products authentication at 90%+ accuracy on AWS) and at Lighthouse.no. Today I'm an ML Engineer at Mowi, operationalizing AI across the value chain.
I care about systems that are private by design, computationally efficient, and actually deployable. Federated learning, LoRA / PEFT, agentic LLMs, real-time vision. The unifying thread is making intelligent systems work at the edge of constraints.
Trained at top institutions: NTNU (PhD) and Imperial College London (visiting researcher). Production ML at Wrist-Shot, then Lighthouse.no, and now ML engineering at Mowi.
Genuinely enjoying the brilliant environment and people at Mowi. Leading the Blue Revolution.
Grateful for a sharp, collaborative team that made hard problems feel light.
A brilliant chapter; loved the collaboration with such sharp minds.
The best mentors I could have hoped for. Their knowledge, engaging personality, support, and genuine friendship helped me grow tremendously as both a human and a researcher.
A brilliant research environment with ideas floating freely, shaping the next generation of machine learning solutions.
Projects I personally found genuinely interesting to build, and that hold up in production. Each had a deliberate choice behind the model, the license, the deployment surface, and the data flow.
End-to-end computer-vision system enforcing PPE compliance (safety vests, hard hats, harnesses) across industrial environments. Built on YOLO-NAS, chosen deliberately for its permissive license (commercially deployable, unlike YOLOv8/v5 GPL variants), combined with Apple's Depth Pro for monocular depth estimation, so the system reasons about proximity to hazard zones in addition to detecting PPE on bodies. Real-time inference with user-adaptive alerting, deployed end-to-end on AWS EC2. Measurably reduced onsite violations.
Multi-agent system that interprets natural-language requests and resolves them across heterogeneous sources: PDF catalogs, multiple internal databases, and product imagery. A router agent decides on the right retrieval strategy per query (structured SQL, semantic PDF search, or visual lookup), then a synthesis agent fuses the evidence into a grounded answer. Built with LangChain for intent & entity extraction, with conversation memory for multi-turn refinement. Replaces manual lookup across systems with one natural interface.
Novel framework for population-based SHM across bridges. Clustered Federated Learning trains without sharing raw data. Bridges clustered by principal-angle similarity (no domain knowledge required), with adaptive parameter transfer for onboarding new infrastructures cheaply.
Integrated drones with a blockchain safety protocol to enhance vehicular security, deployment efficiency, and communication reliability. Optimized for spectral efficiency in dense urban scenarios. Published in IEEE Trans. Intelligent Transportation Systems.
Work shaped by partnerships across leading academic and industrial research labs.
NBL
TUW
NUST
NTNU
ICL
Eleven peer-reviewed papers across federated learning, structural health monitoring, vehicular networks, and wireless communications. Full list on Google Scholar.
From IEEE conference proceedings to mentorship sessions with the next generation of engineers. I enjoy translating dense research into clear, useful ideas.
With Salman Ijaz Siddiqui & Pierluigi Salvo Rossi
On topology-aware learning, federated meta-learning, and how intelligence can self-organize at the network edge.
A career-counseling and outlook talk for engineering students navigating the rapidly evolving AI landscape.
Hands-on across the modern ML stack, from raw research code to AWS-deployed systems.
Doctoral training in Norway, master's in Pakistan, plus AWS certifications.
Topology-Aware Machine Learning in IoT Systems
Machine Learning–Based Blockchain Networks for Combating Security Threats in IoTs
Low Power Dynamic Random Access Memory (DRAM)
"Not the best photographer, but I love to capture."
Mountains, fjords, slow mornings with chai, and wandering through the wonders of the world whenever the calendar allows. A handful of frames from a phone camera and a curious eye.
Open to collaborations and research conversations. Fastest way to reach me is email.