About

From code to cloud to AI

I'm Ali Mehdi, an AI Engineer in Lahore, Pakistan. I design production LLM systems—RAG, agents, APIs—and the MLOps stack that keeps them reliable under real traffic.

My career moved from full-stack web development → cloud & Kubernetes platform work → generative AI in production—including ComfyUI pipelines teams use for real marketing deliverables. Scroll the timeline below to see how each phase built the next.

The beginning

WordPress & PHP Developer

Started my career building client websites and custom themes. Learned how to ship for real users—performance, plugins, and maintainable PHP.

  • WordPress, jQuery, JavaScript front ends
  • PHP backends and theme customization
WordPressPHPjQuery
Expanding stack

Full-Stack Web Developer

Moved beyond CMS work into APIs and modern frameworks. Collaborated with international clients on web products end to end.

  • Laravel applications and REST APIs
  • MongoDB, testing, and deployment workflows
LaravelMongoDBJavaScriptAPIs
Certification Cloud foundations

AWS Developer Associate

Earned my first cloud certification and began architecting scalable backends on AWS for production workloads.

  • EC2, RDS, S3, CloudFormation
  • High-availability patterns for client platforms
AWSEC2RDSS3
Cloud era

Cloud & Backend Engineer

Deepened AWS expertise with enterprise clients—designing infrastructure that scales before traffic spikes, not after.

  • EFS, EBS, multi-AZ database design
  • Infrastructure automation and cost-aware architecture
AWSCloudFormationEFSArchitecture
Platform engineering

DevOps & Kubernetes Engineer

Owned cluster operations, IaC, and observability. Built the platform skills that later became my MLOps foundation.

  • Kubernetes deployments, Terraform, GitOps
  • Prometheus, Grafana, OpenTelemetry pipelines
  • CKAD (2021) and CKA certifications
KubernetesTerraformPrometheusGrafana
CKAD Kubernetes depth

Certified Kubernetes Application Developer

Validated hands-on skills designing and operating workloads on Kubernetes—deployments, services, networking, and troubleshooting.

CKADK8sHelm
Pivot AI engineering

Generative AI in Production

Shifted focus from pure infrastructure to shipping LLM products—RAG, APIs, evaluation, and model deployment.

  • Enterprise RAG over internal docs (LangChain, Pinecone, FastAPI)
  • Streaming LLM APIs with Redis session memory
  • MLflow experiment tracking and SageMaker fine-tuning
  • ComfyUI workflows for repeatable image generation (see 2024)
LangChainOpenAIRAGFastAPIMLflow
Real-world Generative media

ComfyUI Production Pipelines

A marketing team was spending days in Photoshop creating product hero images and social crops for each SKU. I built ComfyUI node workflows that take a plain product photo and output on-brand backgrounds, lighting, and sizes—then wired them to a FastAPI job queue and GPU worker so the team could batch hundreds of assets overnight to S3/CDN.

  • ControlNet + IP-Adapter for consistent brand look across variants
  • ComfyUI API automation—no manual clicking through the UI for production runs
  • Replaced ~3 days/week of designer batch work with a repeatable pipeline
ComfyUIStable DiffusionControlNetFastAPIGPUS3
Present Today

AI Engineer · LLMs · MLOps · ComfyUI

Building agent workflows, vector search, ComfyUI image pipelines, and inference platforms on EKS—with observability, guardrails, and cost controls baked in.

  • Agents & tool-calling with LangGraph and structured outputs
  • ComfyUI + LLM stacks: text copilots and generative visuals in one platform
  • LLM observability: LangSmith, Phoenix, Grafana dashboards
  • GPU inference on EKS, Bedrock, and SageMaker endpoints
LangGraphComfyUIPineconeEKSLangSmith

Let's connect

I write about RAG, agents, and MLOps on the blog. Open to conversations on production AI—reach out anytime.

Email me GitHub LinkedIn