Skip to content

Stack · 4 layers

What I reach for.

Python and TypeScript most days. The model is rarely the hard part; retrieval, orchestration, infrastructure and evaluation usually are.

01 / 04

Generative AI & agents

  • Multi-agent orchestration: LangChain, LangGraph, AutoGen, CrewAI
  • RAG and vector databases: Pinecone, ChromaDB, Qdrant, pgvector
  • Evaluation: Ragas, DeepEval
  • Prompt engineering: chain-of-thought, few-shot, ReAct
  • Fine-tuning with LoRA; local models with Ollama and vLLM
  • Model APIs: Azure OpenAI, AWS Bedrock, Anthropic, Hugging Face

02 / 04

Backend & data

  • Python: FastAPI, AsyncIO, Celery, Django, Streamlit
  • Node.js, Express and TypeScript
  • PostgreSQL, MongoDB, Redis
  • REST, GraphQL and WebSocket APIs
  • Workflow automation: n8n, Zapier, Make

03 / 04

Cloud & MLOps

  • Docker and Kubernetes
  • AWS: Bedrock, SageMaker, EC2, S3
  • GCP: Vertex AI, Cloud Run
  • CI/CD: GitHub Actions, GitLab CI, Jenkins
  • Linux and shell scripting

04 / 04

Front end & design

  • TypeScript, React and Next.js
  • Tailwind CSS and design systems
  • WebGL, Three.js and GSAP
  • Data visualization: Plotly, Streamlit, GeoPandas
  • Figma, Adobe Creative Cloud and Blender

See these tools used in real projects.

View work→