Asfand Yar

Designed and deployed two voice-based AI agents, automating customer interaction workflows for enhanced efficiency.

Developed a post-session feedback bot that autonomously contacted users to collect reviews, sentiment, and session ratings, improving service insights and retention.

Built a proactive membership renewal voice agent that notified users of expiring memberships and guided them through the renewal process.

Integrated speech recognition, conversational logic, and CRM systems to ensure seamless and intelligent customer communication.

Reduced manual follow-up efforts by 53%, saving significant time and improving overall customer engagement rates.

Developed a comprehensive FedX-GAN system for anomaly detection on CICIDS2017/2018 datasets, significantly enhancing system robustness through synthetic data generation.

Integrated encoder-LSTM pipelines to capture temporal patterns in network traffic, achieving approximately 97% client-side classification accuracy while maintaining strict data privacy protocols.

Implemented efficient, privacy-preserving training methodologies across distributed nodes, resulting in improved communication efficiency and reduced latency during model training.

Led the development of a Vertical Federated Learning (VFL) framework for cyberattack detection, successfully processing data from 20 heterogeneous clients while preserving data privacy across partitions.

Implemented client-side Variational Autoencoders (VAEs) and Attention-based Convolutional Neural Networks (CNNs) to extract rich local features without exposing sensitive raw input data.

Designed and optimized a server-side Transformer-based fusion model and multiclass classifier capable of detecting over 17 distinct cyberattack types with approximately 98% accuracy.

Conducted comprehensive performance validation using precision, recall, and F1-score metrics, while simultaneously improving communication efficiency over multiple federated rounds.

Designed and implemented an AI-powered facial recognition attendance system, increasing access control security by 40% and reducing manual processing time by 75%.

Created an innovative machine learning-based sign language detection system using OpenCV and CNNs, improving accessibility for hearing-impaired users.

Optimized image processing models for medical image analysis, resulting in a 30% increase in detection speed without compromising accuracy, enabling faster diagnosis in clinical settings.

Led a cross-functional team of AI engineers to successfully deliver proof-of-concept solutions for real-time object detection using YOLO architecture, subsequently adopted for production use.

Developed and deployed sophisticated computer vision models for industrial automation and surveillance applications, reducing manual monitoring requirements by 60% and improving detection accuracy by 25%.

Built advanced NLP-based sentiment analysis tools for social media monitoring that increased accuracy by 15% over baseline models, providing more reliable insights for marketing teams.

Successfully deployed multiple machine learning models via Flask APIs, creating standardized interfaces that simplified integration into web-based platforms and reduced implementation time by 40%.

Optimized deep learning architectures to reduce model size by 35% and inference time by 28%, enabling deployment on resource-constrained edge devices.

Conducted extensive research and development on AI-based video editing tools, resulting in automated content generation workflows that reduced production time by 50%.

Implemented sophisticated AI pipelines for sentiment analysis on social media data, improving insight extraction speed by 40% and enabling real-time trend analysis for marketing campaigns.

Contributed to the development of EEG-based brain signal analysis systems for healthcare diagnostics using advanced machine learning models, helping to identify patterns associated with neurological conditions with 85% accuracy.