Muhammad Umair
AI Researcher | Consultant | Engineer
Specializing in Generative AI and multimodal systems for enterprises, development organizations, and startups.
Personal Summary
As an AI researcher, engineer, and consultant, I specialize in building and deploying Generative AI and multimodal systems for enterprises, development organizations, and startups.
I have led and delivered over 50 AI and generative AI projects, including large-scale NLP, computer vision, agentic, and multimodal applications. My industry portfolio includes roles as an AI Engineer and consultant for organizations such as ESGnie (Singapore), VikkAI, Lincware, BlueOcean, and others.
With a strong academic and applied foundation in AI, I also serve as an AI Trainer at atomcamp, where I have designed and delivered professional programs on Generative AI, Agentic AI, NLP, and MLOps.
Qualifications
Education
-
Ph.D. 2020-2025
University of Electronic Science and Technology China
Research Domain: Multimodel, Reidentification
-
MS (Computer Science - 2019)
National Textile University, Faisalabad, Pakistan
-
MSc (Computer Science - 2017)
University of Agriculture, Faisalabad, Pakistan
Current Roles
-
Generative AI Consultant
Staff, VikkAI, Superchat, UNDP
-
AI Trainer
AtomCamp
Professional Experience
AI Engineer/Consultant
Atomcamp's AI LMS and myAIpath App
- Conducted professional training sessions on Generative AI/NLP/MLOPs at ADA University in Baku, Azerbaijan
- Led project for Legal Aid Society Pakistan analyzing public sentiment regarding religious and social events
AI Engineer
OrbLogic
- Developed NLP pipelines for interview chatbot application
- Created Automatic QA system within the chatbot
Data Scientist
ESGnie (Singapore) (June 2022 to March 2023)
- Built NLP Models for Q/A, Summarization
- Built pipelines from Data to Model
- Deployed on GCP, Hugging Face, & OpenAI
Projects
Generative AI Projects
- GPT for contextual writing for email marketing
- GPT-powered documents chat application
- AutoGPT-based plugin for live search
- GPT-powered SEO-optimized content generation
NLP Projects
- Custom word-embedding clustering system
- Fine-tuned NER, POS tagger models
- Urdu language sentiment analysis
- Offensive language detection in Roman Urdu
Computer Vision
- Fine-tuned YOLO-5 for custom object detection
- ViT for facial emotion recognition
- Multimodal benchmarks for re-identification
Research Projects
- Content mining of YouTube videos
- Classification of Urdu text
- Malware Classification using ML
- Text Summarization model for talk shows
Technical Skills
Programming
- Python
- JavaScript
- SQL
Tools/Platforms
- Huggingface
- Docker
- GCP
- Livekit
Libraries
- Transformers
- Langchain
- PyTorch
- TensorFlow
Publications
KAT-ReID
Assessing the Viability of Kolmogorov–Arnold Transformers in Object Re-Identification
Content-Based Venue Recommender
Approach for Publication
Detection of bowler's strong and weak areas
In cricket through commentary