I'm an AI Platform Engineer focused on the infrastructure layer that production AI runs on.
My background is in DevOps, Kubernetes, cloud infrastructure, and MLOps. I now apply that systems foundation to LLM serving, reliability, observability, performance, capacity, scaling, and infrastructure cost.
I’m particularly interested in AI-workload-aware infrastructure engineering — understanding how latency, throughput, queueing, batching, scaling, failure behavior, and model-serving patterns should influence platform architecture.
I build public engineering proofs around production AI systems, with emphasis on runtime behavior, failure modes, observability, and measurable trade-offs.
Current work includes:
- ⚙️ LLM serving & inference — latency, throughput, queueing, capacity, replica behavior, scaling, and reliability
- 🧱 AI infrastructure on Kubernetes — deployment, scheduling, platform automation, GitOps, Terraform, and cloud infrastructure
- 📊 Observability & reliability — Prometheus, Grafana, alerting, failure injection, runbooks, and evidence-driven debugging
- 💰 Performance & cost efficiency — understanding how serving and infrastructure decisions affect capacity and cost
Production-oriented ML infrastructure built and validated on live AWS EKS.
Evidence:
- 65 AWS resources provisioned and cleanly destroyed with Terraform
- 5/5 ML pipeline stages validated on Kubernetes
- MLflow + PostgreSQL + S3-based experiment and artifact management
- Prometheus/Grafana observability and alerting
- Failure injection across dataset outage, MLflow outage, and OOMKilled scenarios
- Zero static AWS workload credentials using EKS Pod Identity
A Kubernetes-based LLM inference platform for studying latency, throughput, capacity, scaling, reliability, observability, and cost behavior under measurable workloads.
The goal is to make serving behavior observable and reproducible rather than treating inference as a black box.
👉 InferOps
My projects generally follow this loop:
Expose a production problem → Measure it → Trace the cause across the stack → Change the architecture/configuration → Measure the result
I care about evidence such as:
TTFT · latency · throughput · queueing · saturation · utilization · failure behavior · recovery · cost · before/after benchmarks
- 🛠️ 3+ years working with production infrastructure, Kubernetes, CI/CD, cloud, and DevOps/MLOps systems
- 🎓 10 years teaching Computer Science and Software Engineering
- ☁️ Experience with Kubernetes, Terraform, GitLab CI/CD, GitHub Actions, ArgoCD, AWS, Azure, MLflow, Prometheus, and Grafana
LLM serving · inference performance · Kubernetes AI infrastructure · workload-aware routing · reliability · observability · cost efficiency
My technical progression:
Cloud / DevOps → Kubernetes / Platform Engineering → MLOps → Production AI Infrastructure → LLM Inference Systems
📫 asadhanif3188@gmail.com
🌐 https://asadhanif.com/
🔗 https://www.linkedin.com/in/asadhanif3188