LLM Architecture · Security & Optimization · Agent Design · MCP Orchestration
Senior Machine Learning Engineer at Dayforce — building full-stack ML systems and production pipelines that turn applied ML into scalable, real-world products.
Formerly AI Security Researcher (Sr SE) @ Huawei R&D


I am currently a Senior Machine Learning Engineer at Dayforce, where I design and build full-stack, production-grade ML systems. My work involves developing everything from front-end components in React to back-end data pipelines with PySpark and Delta Tables, with a focus on creating robust, scalable, and fair AI-driven solutions. I manage the end-to-end ML lifecycle, including model optimization, bias mitigation, and MLOps.
Previously, as an AI Security Researcher (level: Senior Software Engineer) at Huawei R&D's Anshi Lab (Burnaby, BC), I worked at the intersection of Malware Analysis and Adversarial Machine Learning, balancing the offensive understanding of malware (AI4Sec) with the defensive hardening of LLMs (Sec4AI). I built autonomous agent ecosystems designed to uncover zero-day threats across GPU, NPU, and edge devices. My expertise is in LLM architecture optimization, Transfer Learning, Multi-agent and Tool Orchestration (MCP), and hardware-aware scaling across diverse physical environments.
At Huawei, I led the architecture and development of an Agentic LLM ecosystem that utilizes multiple agents and tools for zero-day binary malware analysis. My engineering philosophy revolves around pushing the boundaries of AI infrastructure - whether by recasting transformer architectures with Vector Symbolic Architecture (VSA), implementing Mixture-of-Experts (MoE) routing, or optimizing hardware deployment across NPU, GPU clusters, and IoT edge devices to achieve massive production performance gains.
I hold an MSc in Computer Science from the University of British Columbia. My thesis focused on the frontier of AI and software engineering, where my goal was to make software engineering more accessible by exploring how LLMs could reason about code. Before my current role at Huawei, I worked as a Firmware Engineer at MineSense Technologies and as a Research Assistant at the SOftware Analytics and Research (SOAR) Lab at Singapore Management University. I received my BSc in Computer Science and Engineering from North South University, graduating as the Salutatorian with VC's Gold Medal.
I've always gravitated toward the edges of complexity—places where systems struggle with what humans can reason about intuitively. I began my professional journey in academia as a lecturer in my home country, Bangladesh, delivering hundreds of CS lectures and helping students build confidence in subjects that often feel intimidating. That foundation in teaching shaped my core philosophy: break problems down, understand the system, then push it to scale.
Beyond my engineering work, I am an active published researcher with over 150 citations across A* and A-rated conferences. I also remain deeply committed to community service and mentorship. I actively mentor teams building AI tools for social impact, such as speech-to-text LLMs for medical history collection, and have designed programming workshops to empower the next generation of tech leaders.
MSc in Computer Science, 2022
University of British Columbia
BSc in Computer Science & Enginnering, 2018
North South University
Research & Engineering Impact
Operating at the intersection of Malware analysis (LLM for Security) and Adversarial Machine Learning (Security for LLM).
Responsibilities:
Department of Computer Science,
School of Engineering and Physical Sciences
Courses Taught:
Responsibilities:
Department of English and Modern Language,
School of Humanities & Social Sciences
Courses Taught:
& Selected Scholarships
95%
95%
90%
90%
95%
90%
95%
90%
85%
85%
90%
90%
95%
70%
65%
65%
70%
80%
40%
30%
25%
90%
75%
Full Papers in Top Tier Conferences
We investigate the benefits of PTMs for app review classification compared to the existing models, as well as the transferability of PTMs in multiple settings.
In this paper, we propose Adaptive Online Biterm Topic Model (AOBTM) to model topics in short texts adaptively. AOBTM alleviates the sparsity problem in short-texts and considers the statistical-data for an optimal number of previous time-slices.
Papers and Pre-prints
A GPU-accelerated Access Control List (ACL) pattern matching engine achieving 100M packets/second throughput at 1K rules, sustaining ~80K packets/s at 5M rules.
Developed embedded C firmware for a SLAM-based multi-robot communication and motion control suite, including navigation, obstacle avoidance, and robot-to-robot communication protocols.
Optimized a CNN-based Malware Filter Framework (MFF) by implementing Atrous Spatial Pyramid Pooling (dilation), achieving a 315% performance boost through feature-profiling, memory caching, and multi-view filtering.
Lead architect of an Agentic LLM ecosystem with 15+ agents and tools for zero-day binary malware analysis, deployed across NPU, GPU clusters, and IoT edge hardware.
Evaluating Pre-Trained Models for User Feedback Analysis in Software Engineering - A Study on Classification of App-Reviews
AOBTM - Adaptive Online Biterm Topic Modeling for Version Sensitive Short-texts Analysis
The interactive visualization tool makes it easier for the developers to traverse through the extensive result set generated by the text classification and topic modeling algorithms. It also helps developers to quickly comprehend the outcomes of implemented model feature combinations.
The developed tool contains data visualization, trend analysis, and prediction components. The visualization enables the users to interact with the election data through various techniques, including Geospatial visualization.