Ibtisam M

Architected and shipped an AI-powered campaign management platform where an LLM autonomously generates schedule, audience, content, and tone from a single description; integrated Twilio, SendGrid, background queues, CRM writeback — cutting campaign creation time to minutes. Designed and deployed an automated SMS-based class rescheduling system that messaged students who missed classes and redirected them to an AI chatbot interface for instant self-service rescheduling and Q&A.

Developed a full-stack Microsoft Project add-in for natural-language schedule analysis and project-context-aware chat. Architected a real-time security microservice (FastAPI/YOLOv11/DETR) for multi-stream CCTV surveillance. Re-engineered Nginx configuration, reducing API latency from ~25s to <1s. Built a continuous training and evaluation pipeline for object detection models using YOLO, CVAT, MLFlow, and Gradio. Deployed containerized CV pipelines for camera movement detection and PPE-compliance.

Built and deployed a web application for AI-powered outfit visualization and background changing pipeline using Automatic1111 API. Implemented Load Balancers with instance warmup strategies. Fine-tuned Segment Anything Model (SAM) for garment segmentation and RTMPose for full-body pose estimation. Developed a 'Face Swap' web app for BookMyShow, reducing inference latency to ~1 second.

Designed & implemented an MLOps pipeline for a finance firm, ensuring scalable deployment. Managed model versioning with DVC & streamlined dependency management using Poetry & Docker.