I am a Computer Vision Research Engineer at Roll.ai (San Francisco), working remotely. I build monocular systems for 3D reconstruction, depth and motion estimation, and controllable camera motion.
My work lives where perception meets geometry, recovering 3D structure and motion from ordinary video. Earlier, I researched safe multi-robot navigation at Information Technology University and built a license-plate recognition system deployed across seven countries at Hazen.ai.
I am moving toward robotics: motion planning and autonomous systems, carrying the same perception and geometry off the screen and into the physical world.
Research Interests
Roll.ai · 2025–Present
A modular monocular system that coordinates depth, optical flow, segmentation, and reconstruction to recover cinematic camera motion from ordinary video. Segmentation-aware processing suppresses artifacts, and motion-guided ControlNet diffusion enables controllable camera trajectories, yielding stable, high-quality 2K output with a 3× inference speedup.
ITU · B.S. Thesis · 2023–2025
A safety-aware framework for collision-free coordination of 5–50 robots in dense space. A dual-controller design pairs trajectory planning with a fallback safety controller, and a precedence-based decentralized strategy resolves conflicts without a central coordinator. Benchmarked on smoothness, makespan, travel distance, and collision severity.
Hazen.ai · SDAIA · 2024–2025
A recognition system deployed across seven Middle Eastern countries under smog, night, blur, and distortion. A variational autoencoder for OCR, paired with an 84M-sample synthetic-data pipeline, reaches 98% real-time accuracy and generalizes across countries from a single model, with no retraining.
Rawa'a AI · 2025
Deep forecasting across 35+ assets using TiDE and Temporal Fusion Transformers over a 600+ feature dataset fusing market, macroeconomic, and alternative signals. A proposed Trend-Aware Loss jointly optimizes prediction error and directional accuracy.
Turing · 2025–2026
Multi-turn agent-completion workflows spanning tool use, function calling, and real-world reasoning. Modeled assistant behavior across email, calendar, and maps-style tools with correct API-style execution across multi-step tasks.
Intelligent Machine Lab · 2023
Predicting socioeconomic indicators from Sentinel-2 imagery and DHS survey data. Trained CNNs against Vision Transformers over 15K+ images, reaching 76% accuracy with ViTs, alongside an interpretability analysis of poverty-linked feature representations.
Computer Vision Research Engineer, Roll.ai
San Francisco · Remote · May 2025–Present
Undergraduate Research Assistant, Information Technology University
Lahore, Pakistan · Sep 2023–Apr 2025
Associate Computer Vision Engineer, Hazen.ai
Lahore, PK / Saudi Arabia · Jul 2024–Jan 2025
Lahore, Pakistan · CGPA 3.41 / 4.0
Thesis: Safe Navigation of Multi-Agent Robotic Systems
Coursework: Deep Learning, Computer Vision, Artificial Intelligence, Cyber-Physical Systems
Higher Education Commission of Pakistan
Full funding across the entire undergraduate degree.
Ministry of IT & Telecom, Pakistan
Competitive national grant supporting research on scalable multi-robot navigation.
I'm always glad to talk about perception, 3D vision, and robotics, whether it's research, a role, or a hard problem worth solving. Email is the fastest way to reach me.