Product Manager, Lead AI Researcher
BitLogix
Total years of experience :17 years, 6 Months
BitLogix a UK based company leverages over 15 years of industry expertise, specializing in delivering cutting-edge solutions tailored for the healthcare and retail sectors. Its innovative products streamline workflows, enhance patient care, and drive optimal outcomes in the healthcare domain. I am leading research and development in NLP and AI for Localized Speech Recognition, enabling digitization of doctor patient conversation for healthcare domain, focusing on:
• Training Deep Learning Generative AI models using large-scale datasets, optimizing hyperparameters, and employing techniques such as transfer learning and fine-tuning to improve model performance for pre-trained models for instance OpenAI-Whisper.
• Experimenting with different attention mechanisms, such as self-attention mechanisms based on Transformer architecture used in OpenAI-Whisper, to capture long-range dependencies and improve context modelling in speech recognition. Optimizing attention mechanisms to efficiently handle variable-length input sequences and focus on relevant information.
• Evaluating model performance using appropriate metrics and conducting rigorous validation experiments to ensure reliability and accuracy in real-world scenarios.
• Integration and deployment of trained models into production environments, optimized for efficiency and scalability, for real-time speech recognition.
SlimLogix is an AI, ML company to develop high-end technology products and solutions to solve real-world problems. The company is focused around Data Science, Machine Learning, Deep Learning, Computer Vision and Artificial Intelligence based solutions. I have developed scale-able solutions based on contemporary techniques of AI, ML, CV, Deep Learning for Anomaly Detection, Abnormal Human Behaviours Detection, Human Actions Detection, Autonomous Video Surveillance and Image/Video Analysis. I have led AI research to develop an AI based CCTVGuard security surveillance system, solving real-world AI problems focusing on:
• Research and development of Inflated Inception network (I3D) based deep network to learn spatial and temporal information from video frames for real-world anomaly detection in surveillance cameras.
• Research and development of deep Inflated network to learn both spatio-temporal and
motion features from the video frames for real-world anomaly detection.
• Anomalies Detection such as, Abuse, Assault, Fight, Explosion, Road-Accident,
Explosion, Shooting, Shoplifting, Robbery, Stealing.
• Face detection using MTCNN, Face recognition using FaceNet and KNN. Face recognition
and Face liveliness detection in real-time in day and night vision.
• Face recognition and Vehicle number plat recognition to maintain security in the society.
• Fire Detection, Weapon detection using Yolo trained on custom dataset.
• Development of the product for real-time stream processing from Live IP cameras, to
detect anomalies using deep learning models, and generating real-time alerts against
the detected anomalies for the authorities to maintain security.
• Model’s deployments on the cloud for real-time stream processing from IP cameras.
• Development of the application for Person and glass-counting from live cameras.
• Noise removal from the images using Generative models.
• A generic Flask based API implementation on the range of models such as, Anomaly
Detections, Face Recognition, Object Detection, Weapon Detection, Fire Detection.
AnomalyNet: a spatiotemporal motion-aware CNN approach for detecting anomalies in real-world autonomous surveillance 01/2024
UiB - Bergen, Norway, CUI - Lahore, Pakistan
Research paper published in an impact factor journal with the focus on video anomaly detection. The proposed model outperforms state-of-the-art methods achieving high performance in the domain of unsupervised, weakly-supervised, and fully-supervised for the task of anomaly detection in the surveillance videos for real-world autonomous surveillance.
Robust learning for real-world anomalies in surveillance videos 01/2023
UiB - Bergen, Norway, CUI - Lahore, Pakistan
Research paper published in an impact factor journal with the focus on video anomaly detection. The proposed model outperforms state-of-the-art methods achieving high performance for real-world anomalous events recognition for autonomous surveillance.
Fast learning through deep multi-net CNN model for violence recognition in video surveillance 03/2022
UiB - Bergen, Norway, CUI - Lahore, Pakistan
Research paper published in an impact factor journal with the focus on violence detection. The proposed model outperforms state-of-the-art methods in terms of accuracy and has faster learning capability, making it an effective learning architecture on images and videos.
Violence detection in surveillance videos with deep network 12/2018
CUI - Lahore, Pakistan
Research paper focuses on fight detection in videos using deep neural networks.
Games over bluetooth 03/2006
CUI - Lahore, Pakistan
Research paper focuses on writing efficient game development algorithms for resource constraint devices.
SlashNext a US based phishing protection company, features cloud-based real-time phishing detection technology and seamless integration with firewalls and DNS infrastructure. I have led and managed an AI research group of 22+ resources to develop phishing detection technology focusing on:
• AE and Generative models for images similarity, for phishing attacks detections.
• NLP based approaches for Symantec similarity, for phishing attacks detections.
• Data Science for research and analysis of phishing attacks.
• Project planning, Project scheduling, Scrum Management, Agile, meeting moderation, reporting to hire management and goals tracking.
ZoeFit an AI based cognitive wellness app development at Appifytech for foods & activity logging as a personal assistant. Zoe listens and learns every time user use it based on AI.
Development of an innovative cricket platform on web and mobiles enabling the need of modern cricket professionals. I have focused on following:
• Product planning, Product development, Resource planning, Tracking and Ensuring on time delivery of the deliverables.
• Leading team of MEAN and iOS application development.
H2wellness a US based delivers the leading digital technology, thats redefining connected Health and Engagement by providing digital health solutions. I have focused on following:
• USANA True Health Assessment application development for iOS.
• USANA THC, TCMD application development for iOS / Android.
Ciklum is a global digital solutions company for fortune 500 and fast-growing organisations. The company is headquartered in London and has software development centres around the world. I have focused on following:
• Research and development of iOS application.
• Client communications, tasks planning and development.
I have focused on research and development of CRM solution for mobiles.
I have focused on development of CRM and ERP solutions in web and mobiles.
SugarCRM based customised business oriented solution developed for assisting Clients about the company services provides the following key features:
1- PROCESS has three major departments interrelated and working with each other for PRO & IT services.
• Sales: Sales Agents works for marketing companies business. Workflow helps them to manage customer’s data in leads, contracts, Quotations, potential customers as opportunities, and campaigns for marketing.
• Call Center: Call Center Agents are assisting Client’s inquires immediately and directly from system. Workflow helps them to initiate Clients Cases on request, creating Tasks accordingly, sending Cases Quotations to Clients, and Closing the Opened Cases for guaranteeing the quality of service.
• Operations: Operations is responsible for actually doing the Clients requested jobs. Workflow helps them to send Messenger to Clients for documents collection and delivery, archiving these documents within system, sending PRO to government department and ensuring to finish job within timeline.
• Accounts: Accounts department is responsible for managing the Cash collected from Clients to make sure that Client is paying for each and every transaction.
2- Reports: Reports has a central role in Company which helps the Managers for significant decision making. Automatic Reports generation and Auto Emails to Managers are the key features.
3- Provides more than 22 various modules for managing data.
4- Integration with SugarMini the Blackberry application, Email server, LDAP, CISCO Call Manager on Linux platform.
5- Automatic Emails grabber, Popup on Clients phone call for displaying Clients details, Database and Code backup policy.
6- A rich Administration Interface.
7- Dynamic layout customization panel for customizing the modules screen.
8 - SugarMini is a wireless plug-in for Blackberry devices which allows over-the-air synchronization for many SugarCRM modules.
I have focused on mobile applications development in JAVA, J2ME and Symbian OS.
Title: Newstouch - newstouch dot com
As JME Software Engineer
Newstouch is a powerful and fast RSS reader for mobile phones for latest news, web logs, sports, weather updates, stock rates & more. The application uses a remote server to process and compress feeds before sending them to user phone. Only new articles are downloaded to save bandwidth.
Client application designed using MVC architecture and different design pattern like Singleton pattern for different screen controllers and Record store Management System, Factory pattern for canvas creation for supporting portability for different device range that have Nokia Full canvas or MIDP canvas class. The application can download up to 25 items for a single feed and store up to 100 items per feed.
Client application provides the following key features:
1. Manage feeds directly from phone.
2. Categorize feeds in tree view.
3. Read full story of article in summary view with paragraph formatting.
4. Read full story in phone’s browser.
5. Send article links to friends via email.
6. Create keyword feeds to get news based on topic of interest.
7. Images are resized to fit phones display.
8. Articles are stored on phone for offline viewing.
Title: MyTravelr - mytravelr dot com
As Senior Symbian Software Engineer
MyTravelr includes Weather, Clock, World Map, Currency Converter and International dialing codes for Symbian Series 60 2nd & 3rd Editions and UIQ 2 & 3. Weather program keeps you connected with updated forecast anytime anywhere. Clock program - powerful world clock and time management application to check time in five world cities at once.
Title: Mobile Messaging Client
As Senior Symbian Software Engineer
A messaging Client for sending and receiving SMS from cell phones, tracking SMS and maintaining history.
