OPEN TO FULL-TIME ROLES
I build intelligent systems at the intersection of machine learning, information retrieval and healthcare.
Edmonton, Alberta 🇨🇦
Trained as an engineer, currently a researcher, consistently happiest when a model ends up in someone's hands.
I am a Master's student in Computing Science at the University of Alberta, specialising in multimedia and machine learning. Before Edmonton I spent nearly three years at MyMedicalHUB in Dhaka as a software engineer working across AI and iOS — the kind of role where you train the model on Monday and ship it to real users on Friday.
That mix is still how I like to work. I am equally comfortable reading a retrieval paper and profiling a CoreML inference pipeline, and most of what I have built lives somewhere between the two: on-device pose estimation for musculoskeletal assessment, LLM pipelines that draft patient notification pathways for clinicians to review, and a benchmark for entity-oriented retrieval that is currently being written up as a short paper.
What I care about is the distance between a promising result and something a person can actually use. Healthcare is where I have spent most of that effort, but the instinct travels.
MSc in Computing Science (Course-based), Multimedia
Sep 2025 — Present · Edmonton, Alberta
BSc in Computer Science & Engineering
2017 — 2022 · Rajshahi, Bangladesh
Azmal Awasaf
AI Engineer & ML Researcher
Edmonton, Alberta 🇨🇦
HackerEarth Machine Learning Challenge
Top 8% (172 of 2,124) — predicting wind turbine power output from operational parameters.
Telenor Youth Forum
Semi-finalist, top 20 — pitched an app-based counselling platform for at-risk youth.
Startup Istanbul
Global top 200 — IoT smart costume concept for activity tracking and safety.
HULT Prize at RUET
Coordinator of participants — ran team onboarding and programme communications.
The work that matters most, first — production systems at Ameya Health, a retrieval benchmark with Missouri S&T, and two apps on the App Store. Everything else is one click below.
Three years of shipping machine learning that people outside the team actually use.
Ameya Health
May 2026 — Present · Edmonton, Alberta
Ameya Health
Nov 2025 — Dec 2025 · Edmonton, Alberta
MyMedicalHUB International Ltd.
Nov 2022 — Aug 2025 · Dhaka, Bangladesh
Retrieval, autoregressive models and the gap between a benchmark number and something you can trust.
Entity linkers over-annotate. Robust-EO measures how much document-side entity noise LLM centrality filtering removes, whether constrained expansion recovers genuinely missing entities, and whether any of it moves retrieval effectiveness.
Course research project
A patch-based local autoregressive framework pairing a Vector Quantised Autoencoder with a causal convolutional sequence-to-sequence model, and an analysis of where the architecture stops generalising.
Course research project
A local autoregressive latent transition framework with residual modelling and a gated drift mechanism for polyphonic generation on JSB Chorales.
Undergraduate thesis, RUET
A real-time small-object detection benchmark across three YOLO generations, trained on a weather-augmented Bosch Traffic Light Dataset.
Two products carrying models I trained, in the hands of people who never think about the model.
Efficient Musculoskeletal Management Assistant
Turns a phone camera into a movement assessment tool. Guided movements, on-device pose estimation, an NLP symptom triage bot, and a WebRTC channel for live sessions with a clinician.
Software Engineer — ML pipeline, assessment flows, telemedicine
AI-powered fitness and performance assistant
A full-body movement screen that runs from a phone — no wearables — surfacing asymmetry and movement inefficiency, with injury-risk classification driving personalised exercise recommendations.
Software Engineer — pose estimation, injury-risk models, live feedback
Listed because recruiters search for them — but the grouping is honest about where I actually spend my time.
Not the headline — but it is where a lot of the energy for the headline comes from.
Strength & consistency
Lifting is the closest thing I have to a control experiment on myself — one variable at a time, measured over months. It is also why movement-assessment software never felt abstract to me.
Bangladesh → Canada, and onward
Moving from Dhaka to Edmonton reset what I think of as ordinary. I take every chance to get out into the Rockies, and I am slowly working through a list of places that look nothing like either home.
Cooking, cameras, long walks
Cooking for people, taking more photos than I ever edit, and the kind of long unstructured walk where a problem I have been stuck on for a week quietly resolves itself.
Photos landing soon — the layout is already waiting for them.
I'm open to full-time roles — and always happy to talk about retrieval, on-device ML, or anything at the messy edge where research meets a real product.
Looking for
