Saad Ahmad

AI / ML engineer in training.

I go by Ayyzenn. I’m a student at FAST NUCES. I work on machine learning systems that have to hold up outside the notebook — especially anomaly detection, deep learning, and the infrastructure those systems depend on.

I use Arch Linux with i3. Most of my work starts in the terminal: design, break, fix, then write down what survived.

What I Do

AI and machine learning come first. Systems work is what keeps those models honest.

AI & Machine Learning
I work across classical ML and deep learning — SVM, KNN, XGBoost, ANN, CNN, RNN, LSTM, transformers — and applied pipelines for RAG, multi-agent workflows, and time-series anomaly detection. Where it makes sense, I keep inference and retrieval local.

Systems & Infrastructure
I automate with Ansible, ship CI/CD with Jenkins and Docker, and deploy on Kubernetes. For observability and load, I’ve used Prometheus, Grafana, InfluxDB, Telegraf, and Locust. The point is simple: if a model or service runs in a cluster, it should be measurable.

The interesting work sits where these meet — detection pipelines in live environments, agents grounded in real data, and infrastructure that can be reproduced.

Selected Work

Full write-ups are in the posts. A few projects matter enough to list even when they are not posts yet.

On the blog

  • Multi-Agent RAG — Router and generator agents with ChromaDB, HyDE, and query decomposition.
  • Text-to-SQL — Natural-language questions over a database with Ollama, LangChain, and SQLAlchemy.
  • Local RAG Knowledge Base — Offline retrieval with Ollama, ChromaDB, and sentence-transformers.
  • YouTube Summarizer — LangGraph and Gemini for fetch, summarize, and Q&A.
  • Jenkins + Docker CI/CD — GitHub to image build to running container.
  • Ansible for Docker & Kubernetes — Playbooks to prepare Ubuntu servers as Kubernetes nodes.
  • Scraping + REST API — FastAPI/Flask with BeautifulSoup, Selenium, PostgreSQL, and MongoDB.

Also worked on

  • Final Year Project — Real-time anomaly detection in a Kubernetes cluster — Detection over cluster metrics and time series, including statistical sliding-window methods and models such as ARIMA, LSTM, SVM, One-Class SVM, KNN, and XGBoost, with a three-way decision formulation for uncertain cases.
  • Time-series anomaly detection — Broader experimentation on benchmarks such as Yahoo S5 / A1 / A4, supervised and unsupervised setups, scaling, and evaluation.
  • Cloud-native monitoring & load testing — Sock Shop on Minikube, Prometheus/Grafana-style observability, and Locust load tests.

How I Work

I learn by shipping. I prefer real machines, real configs, and real failure modes. Then I write it down so the next pass is cleaner.

If it does not run on my hardware, it does not make the blog.

Browse by focus: DevOps, AI-ML, or Others.

Stack

Linux & Systems

Arch Linux Arch LinuxUbuntu UbuntuGentoo GentooDebian DebianLinux Linux
Bash BashVim VimSSH SSHUnix UnixGCC GCC

DevOps & Cloud

Docker DockerKubernetes KubernetesJenkins JenkinsAnsible AnsibleTerraform Terraform
Azure AzureAWS AWSGit GitGitHub GitHubGrafana Grafana

AI / ML & Data

Python PythonJupyter JupyterPandas PandasNumPy NumPyKeras Keras
MongoDB MongoDBPostgreSQL PostgreSQLSelenium SeleniumKaggle KaggleAnaconda Anaconda

Web & Development

FastAPI FastAPIFlask FlaskDjango DjangoHTML5 HTML5JavaScript JavaScript
Hugo HugoVS Code VS CodeMarkdown MarkdownYAML YAMLC C

Timeline

WhenWhat
2020Started this site and began writing in public.
2021Dual-boot Ubuntu. SSH, Git, Azure — first real infrastructure work.
2022Deep DevOps: Docker, Puppet, Chef, services, Gentoo from source.
2023–2024Final Year Project: real-time anomaly detection in a Kubernetes cluster using a three-way decision approach and ML-based detectors.
2025Applied AI in public: local LLMs, RAG, agentic workflows, Text-to-SQL.
2026AI/ML-first focus — deep learning, anomaly detection, CI/CD, Ansible, Kubernetes, multi-agent systems.

Connect

“They call us dreamers… But we are the ones who don’t sleep.”