MUHAMMAD AHMED GHANI
Lead Machine Learning Engineer
and Researcher at ImagineArt
About Muhammad Ahmed Ghani
Muhammad Ahmed Ghani is Lead Machine Learning Engineer and Researcher at ImagineArt, formerly Vyro AI, currently working in Islamabad, Pakistan. He leads the company's ML team and led research for ImagineArt 2.0 and ImagineArt 2.0 Edit, and built ImagineArt 1.5 Pro.
The work spans text-to-image generation, image-grounded editing, native 4K output, video, speech and agentic systems: model architecture, training, evaluation and the inference infrastructure that serves those models in production. As of , systems he has led are used by more than two million people.
Born on , he studied computer science at the University of Central Punjab in Lahore. Before ImagineArt came five years across Ekkel AI, Kodezi and Sigmetec — leading a team of five through more than twenty projects in speech processing, natural language processing and computer vision, and building the MLOps foundations for AI developer tooling as a founding engineer.
What does Muhammad Ahmed Ghani work on?
Generative models, end to end. Diffusion and flow-matching image models, video generation with temporal consistency, speech systems, and agentic AI — from first experiment through training, evaluation and production inference.
Who leads the machine learning team at ImagineArt?
Muhammad Ahmed Ghani has led the ML team at ImagineArt since March 2024, a seven-person group of machine learning engineers and researchers covering model research, product delivery and production infrastructure. See the ImagineArt ML team page.
Who built the ImagineArt models?
Muhammad Ahmed Ghani led the ImagineArt model programme and built the models together with Muawiz Umar, who worked alongside him throughout, supported by the wider ML team.
Who led research for ImagineArt 2.0?
Muhammad Ahmed Ghani led research and the ML team behind ImagineArt 2.0. These are team-built systems; he separately built ImagineArt 1.5 Pro as sole developer.
Where is he based, and is he available?
Working in Islamabad, Pakistan, and originally from Lahore; working with teams globally. Open to research collaboration and speaking, at muhammad.ahmed@imagine.art or i.am.a.pakistani.programmer@gmail.com.
Models &
Products
Research taken through
to shipped products.
Lead Research · Generative AI
ImagineArt 2.0 & 2.0 Edit
Led research and an ML team, together with Muawiz Umar, across the flagship text-to-image and image-editing models: photorealism, prompt understanding, typography, multi-image composition and identity preservation.
Model Research · Native 4K
ImagineArt 1.5 Pro
Sole developer of the realism-focused successor to 1.5, with native 4K generation, stronger anatomy and prompt adherence, and a 40% faster inference pipeline.
MoE Diffusion · Model Programme
ImagineArt 1.5
Led model development and built the training and evaluation stack across data curation, distributed training and automated evaluation.
Agentic Creative AI
Imagine Assist
Built a personalized control agent that routes image, video and audio tools through one conversational interface with style memory and preference learning.
Model Architecture · Deployment
ImagineArt 1.0
Delivered the first in-house realistic image model, including the MoE architecture, data pipelines, training infrastructure and scalable GPU inference.
Applied AI · Infrastructure
Imagine Platform
Shipped avatar and short-form video systems, then reduced inference latency through quantization, caching and GPU optimization.
The Stack
Tools used day to day.
AI / ML
Languages
Infrastructure
Experience
Where the work has happened.
Lead Machine Learning Engineer and Researcher
- Lead and mentor a seven-person ML team across research, product delivery and production infrastructure.
- Led the ImagineArt model programme, building the models alongside Muawiz Umar.
Sr. AI Engineer / Team Lead
- Led team of 5 engineers, delivering 22+ projects across Speech Processing, NLP, and Computer Vision.
- Managed large-scale deployments on AWS, GCP, and Runpod across global infrastructure.
Founding Machine Learning Engineer
- Founding member responsible for MLOps and infrastructure of Kodezi's core AI products.
- Designed scalable pipelines for VS Code extension integration and real-time code analysis.
Certifications
Courses and certifications completed.
Education
Undergraduate study.
Bachelor of Science in Computer Science
- Lahore, Pakistan · CGPA 3.52/4.00
- Active Lead at Computer Society, organizing international tech competitions.