Building production AI systems

Ali Mehdi

AI Engineer

Lahore, Pakistan

I build production LLM apps—RAG pipelines, AI APIs, agents, and MLOps on AWS & Kubernetes. I also ship ComfyUI image workflows for real teams (e.g. e-commerce marketing batches)—not just demos. Former full-stack and platform engineer.

RAG LangChain ComfyUI OpenAI MLOps FastAPI
About Me Technical Blog

Core Skills

LLMs, RAG, ComfyUI image pipelines, and deployed inference

LLM Apps & RAG Expert
Python / FastAPI Expert
LangChain / LlamaIndex Advanced
Vector DBs & Embeddings Advanced
ComfyUI / Image Gen Advanced
MLOps / Kubernetes Advanced
AWS (SageMaker, Bedrock) Advanced
React / Full-Stack UI Advanced
MLflow / Model Evaluation Proficient

Tools & Platforms

Technologies I use day-to-day on client and internal AI projects

LLMs & APIs

OpenAI (GPT-4o, embeddings) · Anthropic Claude · AWS Bedrock · Hugging Face · Ollama · vLLM

Orchestration

LangChain · LlamaIndex · LangGraph · Celery · Redis · Pydantic · structured outputs & tool calling

Data & Vectors

Pinecone · Chroma · Weaviate · pgvector · OpenSearch · S3 · PostgreSQL · Unstructured.io

ML & Training

PyTorch · scikit-learn · SageMaker · MLflow · W&B · Jupyter · fine-tuning · evaluation harnesses

Generative Media

ComfyUI · Stable Diffusion · ControlNet · IP-Adapter · LoRA · batch API queues · GPU workers · S3/CDN delivery

Deploy & MLOps

FastAPI · Docker · EKS · Terraform · GitHub Actions · GPU nodes · autoscaling inference

Observability

LangSmith · Phoenix (Arize) · Prometheus · Grafana · OpenTelemetry · token & cost tracking

AI Projects

RAG, ComfyUI pipelines, APIs, fine-tuning, and MLOps—with real production use cases

  • All
  • LLM & RAG
  • ComfyUI
  • MLOps
  • AI Products
Enterprise RAG assistant

Enterprise RAG Assistant

LangChain · OpenAI embeddings · Pinecone · FastAPI · React — cited answers over 10k+ internal docs

Document Q&A pipeline

Document Q&A Pipeline

Hugging Face · Chroma · Python — PDF ingestion, chunking, hybrid search, Ragas evaluation

ComfyUI e-commerce marketing pipeline

ComfyUI Marketing Pipeline

Real use case: e-commerce team—SKU photo in → on-brand hero + social crops out. ComfyUI · ControlNet · FastAPI queue · GPU · S3. Replaced days of manual Photoshop batches.

AI-enabled learning platform UI

AI-Assisted Content UI

React · REST APIs — integrated LLM suggestions into an existing learning platform frontend

ML inference on EKS

ML Inference on EKS

Kubernetes · MLflow · S3 — versioned model artifacts, rolling deploys, HPA for inference pods

LLM API gateway

LLM API Gateway

FastAPI · OpenAI · Redis · Prometheus — rate limits, team quotas, streaming SSE responses

LLM observability dashboard

LLM Observability

Grafana · Prometheus · LangSmith — latency, token cost, error rate, and retrieval hit-rate dashboards

SageMaker fine-tuning

Domain Fine-Tuning

SageMaker · Hugging Face · MLflow — smaller model tuned on niche terminology vs. zero-shot GPT

Agent workflow automation

Agent Workflow Automation

LangGraph · OpenAI tools · PostgreSQL — multi-step agent with approvals and audit logging

Geo AI demo

Geo + LLM Demo

React · Leaflet · FastAPI — map UI with natural-language queries over location metadata