Ahmad Mustafa Anis
AI researcher at Carnegie Mellon University studying self-supervised learning and computer vision with Michael Tarr.
About
Building visual systems that learn from unlabeled data.
I'm an M.S. student in AI Engineering at CMU. Previously, I built computer-vision systems at Roll.ai, completed the Fatima Fellowship, and earned my B.S. in Computer Science from IIUI in Islamabad.
I also led Cohere Labs' Asia community and hosted 85+ research sessions.
Research
Self-supervised learning
Robust visual representations that transfer across tasks and domains.
World models
Models that understand environments, plan actions, and predict outcomes.
Experience
-
– Present
Research Assistant
TarrLab, Carnegie Mellon University · Pittsburgh
Sequential multi-objective methods for self-supervised visual representations, with Michael Tarr and Deva Ramanan.
-
– Aug 2026
Deep Learning Engineer
Roll.ai · Remote
Real-time depth and matting for live video bokeh at 2× the speed; active speaker detection that cut event editing from 48 hours to 1; a multimodal agent for automated video editing.
-
– Jun 2025
AI Research Fellow
Fatima Fellowship · Remote
Retrieval-augmented test-time inference for medical vision-language models, reaching 87% on IU-Xray. Advised by Dr. Wei Peng.
-
– Jan 2025
Research Contributor
Data Provenance Initiative, MIT Media Lab · Remote
Audited large-scale LLM training data for shifts in web access, cultural representation and transparency. Papers at NeurIPS 2024 and ICLR 2025.
-
– Feb 2024
AI Fellow
PI School of AI · Rome
One of 10 fellows from 300 applicants; key information extraction from long-context transcripts with LLMs.
-
– Mar 2024
Machine Learning Engineer
Red Buffer · Islamabad
OCR on CPU from 0.6 to 16 FPS for live redaction; CLIP-based logo recognition in cricket broadcasts from 30% to 85%; LLM fact-verification pipelines.
-
– Dec 2021
Deep Learning and Computer Vision Intern
WortelAI · Lahore
Celebrity detection and classification at 88% accuracy; a real-time video search engine with CLIP, Qdrant and Elasticsearch.
News
Invited to review for EvalEval at ACL.
Invited to review for Med Reasoner at CVPR 2026.
Organized the Cohere Labs ML Summer School.
I have also given talks on computer vision and hosted 85+ research sessions for the Cohere Labs community.
Earlier highlights
Completed the Fatima Fellowship, advised by Dr. Wei Peng.
Reviewed for ACL ARR.
Our vision-language model study was accepted at the CVPR 2025 Computer Vision in the Wild workshop.
Our multimodal data provenance work was accepted at ICLR 2025.
Joined the Fatima Fellowship as a Pre-Doctoral AI Research Fellow.
Our AI data commons study was accepted at NeurIPS 2024.
Became NLP Lead for the Bytewise Fellowship.
Selected for the Oxford Machine Learning Summer School.
Joined Roll.ai as a Deep Learning Computer Vision Engineer.
Selected for the PI School of AI with a €12,500 scholarship.
Became ML-Maths Community Lead at Cohere Labs.
Served as Urdu Language Ambassador for Aya, contributing three datasets.
Became Data Science Lead for the Bytewise Fellowship.
Became Asian Community Lead at Cohere Labs.
Graduated from IIUI with a B.S. in Computer Science.
Joined Redbuffer.ai as a Machine Learning Engineer.
Joined Wortel.ai as a Software Engineer.
Joined Wortel.ai as a Deep Learning and Computer Vision Intern.
Began undergraduate studies in Computer Science at IIUI.