Muhammad Taha Salaar

Academic

Career

Verbatim text

Muhammad Taha Salaar
Cell: 923057445566 |  Email: tsalaar.2003@gmail.com
LinkedIn: https://www.linkedin.com/in/tahasalaar/
Address: FLAT 2-C,BUILDING 36,ST18,SECTOR F ,ASKARI-X,LAHORE , Lahore , Pakistan
PROFESSIONAL PROFILE
Software Engineering undergraduate at NUST with strong hands-on experience in AI research, multimodal learning, and
production-grade ML systems. Actively working as a Medical AI Research Engineer, focusing on vision–language models, knowledge graphs, and explainable AI for healthcare applications. Proven ability to bridge theory and practice through research-
driven development, including zero-shot medical image localization, diffusion model research, and graduate-level teaching.
Experienced in building and deploying scalable LLM systems, RAG pipelines, and end-to-end MLOps workflows using modern tools
such as FastAPI, Docker, MLflow, and cloud infrastructure. Strong communicator with a track record of collaborating with clinicians,
teaching advanced concepts, and delivering accessible AI solutions under real-world compute constraints.
EDUCATION
Software Engineering
SEECS , Islamabad , 3.22 (2026)
INTERNSHIP EXPERIENCE
TUKL-DLL Lab (SEECS-NCAI)
20-Jun-2023 - 10-Sep-2023
Began the internship with a strong theoretical foundation by completing Andrew Ng’s Machine Learning and Deep Learning
Specialization, covering supervised learning, neural networks, CNNs, sequence models, and optimization techniques. Transitioned
into independent research on Diffusion Models, focusing on their mathematical foundations, forward–reverse processes, noise
schedules, and training dynamics for generative modeling. Delivered a master’s-level lecture on Diffusion Models, explaining core
concepts, intuition, and practical implementation details to graduate students, bridging theory with real-world applications. Designed
and implemented a lightweight diffusion-based generative model optimized to train efficiently on Kaggle GPU environments, making
advanced generative modeling accessible without high-end compute resources.
Cybersecurity Zone
15-Jun-2025 - 26-Sep-2025
Began the internship with an in-depth research phase focused on Security Operations Centers (SOC) challenges, including alert
overload, false positives, analyst fatigue, and limitations of static SOAR playbooks, supported by a structured literature review of ML-
and NLP-based SOC automation techniques Conducted applied research on AI-Enhanced SOAR architectures, studying ML-based
alert triage, adaptive human-in-the-loop playbooks, and NLP-driven incident summarization to reduce Mean Time to Detect (MTTD)
and Mean Time to Respond (MTTR) Designed a modular, cloud-native SOAR system architecture, including alert ingestion pipelines, ML triage engines, event-driven orchestration layers, and analyst dashboards, emphasizing scalability, explainability, and extensibility
Developed and evaluated machine learning pipelines for alert prioritization, leveraging supervised and clustering-based approaches
to filter false positives, assign risk scores, and support analyst decision-making with confidence estimates Implemented NLP-based
incident intelligence workflows, including automated IOC extraction and concise incident summarization, enabling faster
investigations and clearer handoffs between SOC analysts
FINAL YEAR PROJECT
ARIES - Automated Response and Intelligent Enterprise Security
ARIES revolutionizes SOC operations by integrating ML-driven alert triage, NLP-based summarization, and human-in-the-loop
orchestration. It intelligently automates repetitive tasks, empowers analysts with contextual insights, and delivers adaptive, explainable, and resilient cybersecurity operations — advancing both innovation (SDG 9) and institutional security (SDG 16)

AI enrichment