Muhammad Ali Hassan | AI Engineer

01About Me · Headquarters

About Me

AI engineer and Fulbright Scholar pursuing an MS in Computer Science at the University of Illinois Chicago. I build AI models and the software systems that carry them, designed to grow with their users and stay cheap enough to run anywhere, so they reach the people who need them most.

concurrent users on one commodity server: 100+
less CPU inference time vs. baseline: 93%
the cost of a conventional flood station: 1/10
OJ-CS journal paper, 2026: IEEE

Why I build

In September 2014 I was thirteen, crouched behind sacks of flour stacked against our front door in Multan while the Chenab flooded. The water never reached our neighbourhood. It did reach the slums by the river, and that was the day I saw two parallel cities: one prepared, one left to fend for itself.

That divide still shapes how I build. Technology only helps people if it can grow to reach them and stays cheap enough to keep running, so I care as much about the system as the model. At LUMS that meant a flood monitor built for a tenth of the usual cost, a five-camera wildfire network covering 300 km² of Northern Pakistan, and an LLM and blockchain platform that gives clinicians governed, auditable access to sensitive health data.

At Tibbling Technologies I built a compact neuro-segmentation model and the platform that carries it, with fine-tuning in the browser and real-time 3D segmentation. Moving it from AWS to a single commodity server ended our recurring cloud bill while serving 100+ users at once. Labs at the University of Washington and in Zurich validated it, and the Broad Institute used it for analysis supporting autism drug discovery.

Now in Chicago on a Fulbright, I’m going deeper into scalable AI and distributed systems, and I plan to take them home: affordable, local-language AI for healthcare and education in Pakistan.

Download résumé Academic CV Email me LinkedIn GitHub

02Experience · AI Engineer

Tibbling Technologies

Neuroscience labs shouldn’t need a GPU cluster to segment their images. I built the models, and the platform around them, that make that possible.

faster CPU inference, beating SAM on Dice: 93%
neuro-preclinical images evaluated: 36K+
to adapt to a new imaging domain: 1 patch
concurrent users on 100 GB+ volumes: 100+

AI Engineer

Tibbling Technologies · Neuro-AI startup

Jan 2025 – Aug 2026 Remote · Seattle

Advisor: Dr. Asim Iqbal (Weill Cornell Medicine)

Built a 200K-parameter 1D segmentation model in PyTorch that cuts CPU inference time by 93% while exceeding Meta’s SAM in Dice score across 36,000+ neuro-preclinical images. Manuscript in preparation for Nature Methods.
Generalised it into a universal neuro-segmenter trained on self-supervised datasets of neurons and axons, with CPU-only few-shot fine-tuning (DoRA / FiLM adapters) that adapts to a new imaging domain from one annotated patch.
Shipped a full-stack Next.js app for in-browser fine-tuning and real-time 3D segmentation, validated by labs at the University of Washington and Zurich. The Broad Institute used it for colocalisation analysis supporting autism drug discovery.
Migrated the Dockerised stack from AWS to a single commodity server by moving storage to OME-Zarr and inference to ONNX Runtime. That eliminated recurring cloud costs while scaling to 100+ concurrent users on 100 GB+ 3D medical volumes.

PyTorch
ONNX Runtime
OME-Zarr
Next.js
Docker
AWS
DoRA / FiLM
Self-supervised learning

03Research · Cloud Computing Research Lab, LUMS

Governed AI for Healthcare

Hospitals want to share data; regulators need proof it was shared correctly. We built infrastructure that gives clinicians natural-language analytics, with an audit trail no one can rewrite.

workflow types generated near-perfectly: 17
patient records in survival analyses: 400K+
immutable audit trail: On-chain
published in IEEE OJ-CS: 2026

Research Assistant

Cloud Computing Research Lab, LUMS

Jun 2024 – May 2025 Lahore, Pakistan

Advisors: Dr. Basit Shafiq (LUMS) · Dr. Jaideep Vaidya (Rutgers)

Designed an LLM pipeline that turns natural language into BPMN workflows for survival analysis (Cox models, Kaplan–Meier curves), giving clinicians governed analytics through a MERN app. Structured prompting yielded near-perfect executable workflows across 17 types.
Built a blockchain access-control system (Solidity, Ethereum / Sepolia) that authorises dataset requests against XACML policies and logs an immutable audit trail on-chain, keeping patient data off-chain.
Ran an ablation across four prompt settings on SEER and German Cancer Registry data (400,000+ patients), mapping the trade-offs between prompt richness, workflow detail and latency.

LLMs
Prompt engineering
BPMN
Solidity
Ethereum
XACML
MERN
Survival analysis

Read the IEEE OJ-CS paper

04Field research · Edge AI for disaster response

Wildfire & Flood Monitoring

Conventional monitoring stations cost too much for the remote valleys that need them most. These two systems watch for wildfire and flood across Northern Pakistan on a shoestring budget.

of forest covered with WWF: 300 km²
autonomous PTZ cameras: 5
hotspot localisation accuracy: 99%
the cost of a conventional station: 1/10

Wildfire watch network

IoT Lab, LUMS · with WWF · Research Apprentice

Sep 2024 – Jan 2025 Lahore, Pakistan

Advisor: Dr. Murtaza Taj

Shipped a 5-camera PTZ network with WWF covering 300 km² of Northern Pakistan to catch forest fires early.
Automated pan / tilt / zoom sweeps through the Hikvision API with overlap-aware capture planning, so every ridge gets inspected.
Python / OpenCV + YOLO inference localises hotspots at 99% accuracy and raises early alerts for the authorities.

Python
OpenCV
YOLO
Hikvision API
PTZ control

AIoT flood & wildlife monitor

Centre for Water Informatics & Technology, LUMS · Research Intern

Jun 2024 – Aug 2024 Lahore, Pakistan

Advisor: Dr. Talha Manzoor

Architected an ESP32-CAM monitor for floods and wildlife where conventional stations are too expensive to deploy.
Screens frames on-device with Edge Impulse before a GSM upload to server-side YOLOv5, delivering monitoring at 1/10th the cost.
Used sleep modes and watchdog timers for reliable, power-efficient operation in internet-limited regions.

ESP32-CAM
Edge Impulse
TinyML
YOLOv5
GSM

05Projects · The Garage

Projects

Things I’ve built because I wanted to understand them: consensus protocols, on-device LLMs and real-time platforms.

Raft-based fault-tolerant key-value store

Go · RPC · Raft

Implemented the Raft consensus protocol from scratch: leader election, log replication and crash-recovery persistence.
Validated against a test suite that simulates network partitions, dropped RPCs and node restarts.

Go
Distributed systems
Consensus

On-device LLM profiling

Python · Ollama · ADB · Android

Advisors: Dr. Ihsan Ayyub Qazi · Dr. Zafar Ayyub Qazi

Built a memory-tracing pipeline with Ollama and ADB to profile a 2-bit quantised Llama 3.2 1B across low-cost Android devices.
Identified KV-cache growth and cold-start latency as the primary bottlenecks limiting on-device inference.

LLMs
Quantisation
Android
Profiling

GitHub

Full-stack hiring platform

React · Node.js · Express · MongoDB · Socket.io

Built asynchronous video interviews, real-time recruiter–candidate chat and application tracking from posting through offer.

MERN
WebSockets
Video

Live demo GitHub

More from the garage

Distributed hash tablePython · sockets

Peer-to-peer DHT with fault tolerance for reliable retrieval in decentralised networks.
RAG research chatbotLangChain · Pinecone · FAISS

Answers questions by retrieving from a corpus of research papers and Wikipedia.
Real-time trading app MERN · Socket.io

Instant offer updates and inventory sync for an online marketplace.
Cloud billing systemFastAPI · Nginx · SQL · Oracle Cloud

Three-tier billing web app for bill retrieval, payments and adjustments.
UNIX-style shellC

Command-line shell with pipelines and command chaining.
Battery-less IoT securityResearch · Mementos

Showed that charge/discharge cycles leak which app an intermittent device is running.

06Publications & awards · Hall of Records

Publications & Awards

Peer-reviewed work, and the people who bet on me.

Publications

Governance-Aware AI Workflow Infrastructure for Secure and Policy-Compliant Healthcare Analytics

S. Ali, M. A. Hassan, H. Ahmad, A. Jai, B. Shafiq, et al.

IEEE Open Journal of the Computer Society, vol. 7, pp. 1377–1388, 2026
CPU-efficient neuro-segmentation (Tibbling Technologies)

M. A. Hassan et al.

Manuscript in preparation for Nature Methods

Awards & honours

Fulbright Master’s Scholarship2026
Winner, Techstars Startup Weekend Lahore2025

Most innovative startup, for low-resource-language AI.
Winner, Ideation Academy startup incubator2023
Winner, Future Synergy AI-powered Innovation ChallengeIncubation placement, Azerbaijan
Dean’s Honor List, LUMS2021–22
Merit Scholarship, LUMS2021–22
100% Merit Scholarship, A LevelsBloomfield Hall School
NVIDIA DLI: Fundamentals of Accelerated Data ScienceCertification

07Education & teaching · The Academy

Education

From Lahore to Chicago, and a lot of office hours in between.

MS, Computer Science

University of Illinois Chicago

Sep 2026 – Jun 2028 Chicago, IL

Fulbright Scholar

BS, Computer Science

Lahore University of Management Sciences (LUMS)

Sep 2021 – May 2025 Lahore, Pakistan

Dean’s Honor List · Merit Scholarship

Coursework: Data Structures, Algorithms, Operating Systems, Distributed Systems, Databases, Blockchain, Deep Learning, Machine Learning, Network Security, Topics in LLMs, Internet of Things.
Summer exchange at the University of Baltistan (2023).

Teaching

Head TA, CS 100: Computational Problem SolvingSummer & Fall 2022

Led a team of 4 TAs supporting 200+ students; mentored C++ labs across 6 project groups.
TA, CS/EE 3812: Blockchain Technology & ApplicationsSpring 2023

Ran tutorials on Solidity and contract deployment for 40 grad and undergrad students; wrote assignments, quizzes and exams.
TA, CS 210: Discrete MathematicsFall 2023

Office hours and assignment support on logic, combinatorics, graphs and proofs for 200+ students.

08Leadership & community · The Arena

Leadership

Teams I’ve started, led or grown, from speech technology for under-served languages to a 1,000-person esports scene.

of Pashto speech collected: 10+ hrs
esports event attendees: 1,000+
member team led: 50+
in sponsorships raised: PKR 200K

CEO & Co-Founder

Project Indigenous

Jan 2025 – present

Tackling Pakistan’s speech-tech gap with ASR / TTS for under-served languages. We collected 10+ hours of Pashto speech and won startup incubation.
Pitched at Techstars Startup Weekend Lahore and won most innovative startup.

ASR
TTS
Low-resource NLP

Co-Founder & General Secretary

E-Gaming at LUMS

May 2023 – Jun 2024

Led a 50+ member team running high-profile esports events with 1,000+ attendees and PKR 200K in sponsorships.

Also

Committee Director, LUMS Model United Nations2023–24

Directed a 50-member team across 20+ academic sessions.
Director Media, SPADES & DANCELUMS2021–23

Led media teams of 10+ reaching an audience of 50,000+.
Peer Ambassador for Social Support, LUMS2023–24

Mentored first-year students through academic and personal transitions.

09Skills & toolkit · The Armory

Skills

The utility belt: the tools I reach for, from training models to shipping them to a microcontroller.

Languages

Python
C++
C
Go
TypeScript
JavaScript
Solidity
SQL
Haskell

ML & AI

PyTorch
ONNX Runtime
LangChain
YOLO
OpenCV
NumPy
scikit-learn
Hugging Face
DoRA / LoRA
RAG · FAISS · Pinecone

Systems & infra

Docker
AWS
Linux
Git
Raft / consensus
OME-Zarr
Nginx
Oracle Cloud

Web

React
Next.js
Node.js
Express
MongoDB
FastAPI
REST
Socket.io

Edge & IoT

ESP32
Edge Impulse
Hikvision API
Arduino
GSM

Blockchain

Solidity
Ethereum
XACML
Infura
MetaMask

Science tools

Napari
Fiji

10Contact Me · The Signal

Contact Me

I’m open to research collaborations, internships and conversations about efficient, impact-driven AI. Email is the fastest way to reach me.

Send the signal Résumé (PDF)Academic CV (PDF)