Muhammad Umar

LLMOps Engineer

Lingoscribe.ai • Temps partiel

Dec 2025 - Present • 5 mos

LingoScribe is an AI-powered speech-to-text platform focused on under-served and local languages of Pakistan. The main goal of the project is to provide accurate and reliable transcription for regional languages that are often poorly supported by existing speech recognition solutions. My work on LingoScribe is focused on deploying and operating the speech-to-text infrastructure. I containerized the Whisper-based inference pipeline using Docker to ensure consistent deployments across environments. This made the service easier to manage, update, and scale. I deployed the transcription service on Kubernetes to support concurrent transcription requests and ensure reliable inference serving. Resource configurations were tuned to balance performance and efficient CPU/GPU usage. Unlike Pehchan, LingoScribe is intentionally designed as a focused and lightweight system. It provides speech-to-text functionality only, without complex downstream NLP processing. This simplicity helps keep the system stable and maintainable while delivering accurate transcriptions for local languages. The project is ongoing, with continuous improvements in model accuracy, deployment reliability, and infrastructure scalability.

Devops Engineer

Emaago Tech • Temps plein

Jul 2022 - Present • 3 yrs 10 mos

- Deployed and managed containerized applications using Kubernetes, ensuring scalability, reliability, and efficient resource utilization for cloud-native AI/ML workloads. - Automated infrastructure provisioning with Terraform and AWS CloudFormation, creating reproducible, scalable, and AI/ML-ready cloud environments. - Integrated AI/ML workflows into cloud platforms, optimizing data pipelines and model deployment using services like AWS SageMaker, Azure Machine Learning, and custom Docker containers. - Collaborated with data science teams to streamline machine learning model training, testing, and deployment using automated CI/CD pipelines. - Developed and maintained cloud security policies, ensuring secure data storage and compliance for sensitive AI/ML datasets across AWS and Azure. - Integrated monitoring and alerting solutions, such as Grafana, CloudWatch, and Sentry, providing real-time insights into AI/ML model performance and system health.