Hi, I'm
Sarah Salah
Computer Scientist - AI specialist
Architecting the next generation of intelligent systems, where AI meets infrastructure, and complexity becomes clarity.
01. About Me
Who I Am
I'm an AI Developer & Software Architect designing complex, enterprise-scale systems where intelligent infrastructure meets production reality. I work at the intersection of architecture, machine learning, and operations; building platforms that are smarter, more stable, and self-healing by design.
My focus is designing intelligent systems for complex infrastructure problems, from automated Root Cause Analysis to anomaly detection pipelines that catch failures before they happen. I believe the next frontier isn't just deploying models, it's making infrastructure that thinks.
When I'm not engineering systems, I'm exploring the latest in AI research and open-source projects, contributing to academic research, and studying the mathematics that makes it all possible.
02. Skills
Tech Stack
AI & Machine Learning
AIOps/MLOps & Infrastructure
Software Architecture
Data & Search
DevOps & Cloud
Backend Development
03. Projects
What I've Built
Log Anomaly Detection Pipeline
Production-grade unsupervised anomaly detection system for Kubernetes logs , combining embeddings, Isolation Forest, and a PyTorch Autoencoder with LLaMA-generated explanations.
Legacy-to-Kubernetes Migration Architecture
Architected a strategic migration path for legacy systems to Kubernetes , preserving business continuity by avoiding full rewrites. Designed an incremental containerization approach that moved workloads to K8s efficiently while maintaining backward compatibility and minimizing downtime.
OpenSearch Log Intelligence
Built intelligent log querying and analysis pipelines on OpenSearch , enabling real-time namespace/container filtering, time-range queries, and anomaly surfacing at production scale.
AI Service Catalog
Designed and deployed an internal AI service catalog for on-premise environments, enabling teams to discover, consume, and monitor ML models via a unified API layer.
Reverse-Engineering IBM Watson Assistant
Reverse-engineered IBM Watson Assistant to build an internal, self-hosted conversational AI platform with custom chat flow logic. Replaced the proprietary engine with an open-source stack while preserving dialogue management capabilities.
LLM Self-Hosting on Kubernetes
Deployed and optimized open-weight LLMs (Qwen3 72B, LLaMA) on 4x NVIDIA L40 GPUs using vLLM with tensor parallelism, served via Kubernetes with full observability stack.
04. Blog
Latest Writing
The Art of Reverse Engineering
A deep dive into the methodology of reverse-engineering proprietary systems and rebuilding them as open alternatives ,the process, the challenges, and the lessons learned.
Read more →Staying Sane in the AI Race: How to Actually Keep Up
Practical strategies for navigating the overwhelming pace of AI advancement ,what to read, what to skip, and how to build lasting intuition in a fast-moving field.
Read more →05. Contact
Get In Touch
Available for freelance projects, consulting, and interesting collaborations. Feel free to reach out, I usually respond within 24 hours.