Khizar Hussain
PhD Student & Software Developer
Summary
PhD student in Computer Science at Virginia Tech with 3+ years of industry experience leading ML and systems projects, delivering 50% efficiency gains and $500K savings while managing 5 engineers. Focusing on Machine Learning, Computer Vision, Computer Systems and Operating Systems, with expertise in Python, C++, PyTorch, TensorFlow, and cloud platforms (AWS/GCP).
Education
Virginia Tech
PhD in Computer Science
2024 - 2028 (expected)
GPA: 3.85/4.0
Blacksburg, Virginia
National University of Computer and Emerging Sciences
Bachelor of Science in Computer Science
2017 - 2021
GPA: 3.73/4.0
Lahore, Pakistan
Work Experience
DSPL Lab - Virginia Tech
Research Assistant
2024-09 - Present
Blacksburg, VA, USA
- • Researched and implemented advanced machine learning algorithms to enhance the effectiveness of the hallucination detection framework, resulting in a 30% increase in accuracy for detecting fake information within LLMs.
- • Developing a unique hybrid approach that combines black box and gray box techniques for improved hallucination detection capabilities on both open source and closed source LLMs (ChatGPT, Llama, Claude) for preemptive mitigation.
ROSA Lab - Virginia Tech
Research Assistant
2024-01 - Present
Blacksburg, VA, USA
- • Conducted benchmarking analysis on the moo-kernel idea and published a workshop paper as second author titled 'Eliminating eBPF Tracing Overhead on Untraced Processes' at ACM SIGCOMM 2024 Workshop on eBPF and Kernel Extensions.
- • Researched methodologies to extend push button verified operating systems such as Hyperkernel to overcome the limitations of running verified eBPF extensions in unverified environments such as the Linux kernel.
Virginia Tech - Computer Science Department
Graduate Teaching Assistant
2024-01 - Present
Blacksburg, VA, USA
- • Head TA for CS3214 (Computer Systems) - Spring 2025: Leading a team of TAs, managing course logistics, and mentoring 300+ students in systems programming, concurrency, and performance optimization.
- • TA for CS3214 (Computer Systems) - Fall 2024: Assisted in teaching operating systems concepts including process management, synchronization, memory management, and file systems to 300+ students.
- • TA for CS2505 (Computer Organization) - Spring 2024: Supported students in learning computer architecture, assembly language (x86-64), and low-level programming concepts.
Abyss Solutions
L6 Data Platform Engineer
2023-02 - 2024-03
Remote - New South Wales, Australia
- • Led a team of 5 engineers focused on data ops, computer vision, and ML pipelines. Participated in recruiting to acquire talent in data science and ML, supporting project needs and enhancing AI capabilities.
- • Collaborated with ML teams to successfully launch Graph Assisted Tagging and 2D ML Asset Labeling initiatives, resulting in a 50% increase in operational efficiency and a 75% reduction in manual labor hours.
- • Automated deployments to V2 product through migration tool, reducing deployment time from 3 days to an average of 2 hours, leading to a 60% decrease in time-to-market for new features.
- • Developed CLI tools fabric-deploy-v2 for managing V2 deployments, enabling seamless data purge, export, import, transfer, and transformation tasks across MongoDB Compass compatible bson files and network connections. Increased data processing speed by 40% on average.
App Rocket
Software Engineer
2022-09 - 2022-12
Lahore, Pakistan
- • Built backend API for 3i Capital investment platform, integrating Slack, Stripe, MailChimp, and Docusign
- • Implemented E2E tests with AWS Canaries and Lambda, reducing live-site turnaround time by 60%
- • Designed custom email verification using SMTP servers, sockets, and data streaming
OMNO AI
Computer Vision Research Engineer
2020-08 - 2021-07
Lahore, Pakistan
- • Researched and engineered a realtime ML and CV pipeline using DeepSort with YOLOv5 to localize and identify soccer players in broadcast video streams, achieving over 85% accuracy achieving state of the art fps.
- • Deployed a cutting-edge jersey number detection and recognition system using a fine-tuned YOLOv5 and ResNet CNN, resulting in close to 90% accuracy on both tasks, enhancing the efficiency of player identification during games.
Publications
Eliminating eBPF Tracing Overhead on Untraced Processes
eBPF '24: Proceedings of the ACM SIGCOMM 2024 Workshop on eBPF and Kernel Extensions
2024