M Ali Hamza

AI Research Engineer at INFINIQ, Seoul, South Korea

Also, PhD candidate at NDSLabs

Phone: +82-10-3305-1540

Address: Seoul, South Korea.

News

Jan 2026: Received the Best Employee (AI research engineer) of the year award at INFINIQ.
May 2025: Joined INFINIQ as an AI Research Engineer in Seoul.
Feb 2025: Our paper "Channel Attention for Fire and Smoke Detection" has been published in Sensors (co-first author).
Dec 2024: Joined VisionIn as an AI Research Engineer in Seoul.
Feb 2024: Our paper "Cyber5Gym: 5G Cybersecurity Training" has been published in Electronics.
Jun 2023: Our paper "E-NSSA: Efficient Neural Architecture Search" accepted at Korea Computer Congress (KCC) 2023.

(2025.05–Present) AI research engineer at INFINIQ, Seoul, S. Korea. (On-site)
(2024.12–2025.05) AI Research Engineer at VisionIn, Seoul, S. Korea. (On-site)
(2020.3–2024.10) AI Research Engineer at NDSLabs, Sangmyung University, South Korea. (On-site)
(2020.8–2024.8) AI Researcher at Canvass Labs, Inc. California, USA. (Remote)
(2019.9–2020.3) Software Engineer at Finz Games, Lahore, Pakistan. (On-site)
(2017.2–2019.5) Associate Software Engineer at System Junction Pvt, Bahawalpur, Pakistan. (On-site)

Prof. Hyun-Chul Kim [hkim@smu.ac.kr] (Professor at Sangmyung University, South Korea).

Prof. Akmal Khan [akmal_shahbaz@yahoo.com] (Professor at The Islamia University of Bahawalpur, Pakistan).

Mr. Peter Shin (CEO at Canvass Labs, Inc., California, USA).

(2026) AI CCTV Surveillance (AI-enabled Video Summary Generation).

Developed a computer vision-based video synopsis pipeline to compress hours of CCTV footage into concise, event-focused surveillance summaries.
Built long-video analytics workflows using object detection, multi-object tracking, spatiotemporal tube extraction, event filtering, and summary rendering.
Designed fall-down event search modules that combine tracked-ID association with detection and configurable before/after event windows.
Implemented object collision-aware scheduling and temporal tube placement algorithms
Developed AI algorithms to reduce object overlap, improve visual clarity, and support faster incident review.

(2025) AI CCTV Camera Surveillance (VLM-based).

Trained Open-vision, Florence-2, Clip-vit, and MobileClip2 models for video surveillance.
Trained a vision language model-based fire and smoke detection system using real-world CCTV surveillance.
Built, fine-tuned, and tested detection vision models for safety (people fighting, humans falling down, human intrusion).

(2024 ) Generating Text from Source Code (Code-2-Text).

Developed transformer-based encoder-decoder for natural text generation from code.
Worked on SOTA aglo such as LoRa, PEFT, Bitsandbyte, Transformers, Seq2Seq, and multimodal.
Optimized TopicGPT and BERTopic for semantic analysis of the generated text.

(2024 ) Multimodal Vision Language Models (Llama 3.2 Vision, NVLM, Llava).

Working to detect patterns based on facial expression, audio, and text.
Training multimodal vision language models for processing YouTube videos to detect common patterns.
Fine-tuning Llama 3.2 Vision model on various videos, audio, and text to find patterns.
Hands-on experience on recent vision LLMs (Llama 3.2 vision, NVLM, Llava, Molmo, etc.)

(2024 ) AI Chatbot for Research Assistant using LLaMA 3.2

Developed a cutting-edge AI-powered research assistant chatbot using the LLaMA 3 model.
Fine-tuned LLaMA 3 for natural language understanding and generation tasks, enabling the chatbot to assist researchers by generating accurate research summaries, technical insights, and literature reviews.
Applied Generative AI techniques and Prompt Engineering to enhance the chatbot’s response accuracy and efficiency for academic and technical queries.
Integrated multi-modal features to process various inputs like text, code, and research papers, improving user interaction and experience.

Technical Skills

Large Language Models:

Training, fine-tuning, and deploying LLMs (Qwen-3/2 series, LlaMA 3/2/1 series, BERT, GPT-4, Alpaca, Mistral, Solar, NVLM, Llava, Molmo).
Text Classification, Sequence Labeling, Text Generation, Transformers, Seq2Seq.
LoRa, PEFT, Adapter, PyTorch, DeepSpeed, Bitsandbtye.
Experience with enhancing models like LlaMA, BERT, GPT, Falcon, Mistral, and other transformer-based architectures.

Computer Vision & VLMs:

Image classification, object detection, semantic segmentation, anomaly detection.
OpenCV, YOLO, Faster R-CNN, and SSD for vision tasks.
Training Open-vision, Florence-2, Clip-vit, and MobileClip2 models for video surveillance.

Generative AI:

Generative AI techniques and applications (GANs, Text-to-Image Models, Generative Data Augmentation, Vision LLMs).
Generative models, Diffusion models. Parameter-efficient fine-tuning (PEFT), Retrieval-augmented generation (RAG).

Machine Learning & AI Frameworks

TensorFlow, PyTorch, Scikit-learn, and Keras.
Hugging Face Transformers and Apache Spark.
Attention Mechanisms, Transfer & Reinforcement Learning, Semi-supervised Learning

Data Science & Analytics:

Data cleaning and preprocessing.
Feature Engineering and selection.
Machine Learning Algorithms (Supervised, unsupervised, semi-supervised, reinforcement and transfer learning.).
Model evaluation and tuning (e.g., cross-validation, hyperparameter optimization).
Data visualization and storytelling (using tools like Tableau, Matplotlib).

Software Engineering & DevOps:

MLOps practices for deploying and managing ML models in production.
MLflow, Docker, Kubernetes, GitHub, Jira, Agile, CI/CD Pipelines, Flask, REST APIs.
Building and managing data pipelines and ETL processes.
CI/CD pipelines for automated testing and deployment.
Collaboration and version control best practices.

Tools & Technologies:

Other Key Competencies:

Page updated

Google Sites

Report abuse