Browser walkthrough
Autoplay · Muted · LoopAnimated placeholder — replace it with your real screen recording.
Try live demoSelect a case study
Choose a product. The complete story below updates instantly.
Portfolio / GERMANY / WORLDWIDE
JOAB DAVID
ENGINEERING PRODUCTION AI SYSTEMS THAT SHIP.
OPEN TO ROLES IN MUNICHPRODUCTION RAG ASSISTANT
Answers you can verify.
Grounded retrieval, streaming responses, tool execution, and confidence-aware citations.
Where to next? →
Knowledge workspace
Active users18K chats
Reliability99.9%
9:41● ◒ ▰
PRODUCTION RAG ASSISTANTAnswers you can verify.
Grounded retrieval, streaming responses, tool execution, and confidence-aware citations.
Search ⌕
Knowledge workspace
Curated for you
Featured projectLLM Chatbot
Web appiOS
Product in motion
Don’t take my word for it.
Watch it work.
Short, distraction-free walkthroughs show the same product across web and iPhone. Add your recordings once; the template handles playback everywhere.
iPhone walkthrough
Autoplay · Muted · LoopAnimated placeholder — replace it with your real screen recording.
Try live demoEngineering evidence
Built beyond
the happy path.
Concrete implementation decisions that make the product maintainable, deployable, observable, and fast.
01
Modular programming16 packagesPrompt, retrieval, tools, memory, policy, and evaluation modules expose typed Pydantic contracts.
02
DeploymentMulti-AZECS Fargate, blue-green releases, Terraform, GitHub Actions, and automatic rollback alarms.
03
First token420 msRedis semantic cache, parallel retrieval, streamed OpenAI responses, and connection pooling.
04
Quality gate94.2%DeepEval faithfulness, answer relevance, retrieval precision, and adversarial regression suites.
System architecture
Control flow,
without the mystery.
Each request is authenticated, routed, observed, and returned through explicit boundaries. Hover a node to focus its responsibility.
Live production request LLM Chatbot
- HTTPS
- SSE
- WAF
- Rate limits
- Next.js
- Cognito
- FastAPI
- LangGraph
- SQS
- EventBridge
- Embeddings
- Reranker
- OpenAI API
- Guardrails
- Redis
- TTL
- Postgres
- pgvector
- DeepEval
- Langfuse
- OTel
- CloudWatch
Client Edge AWS Backend Intelligence Data Evaluation
Selected work / 4 projects
Built to be used.
Designed to be remembered.
About me
Machine Learning Engineer · Germany
I build complete AI products - from model experiments to reliable production systems.
I am a full-stack Machine Learning Engineer based in Germany, specializing in LLM systems, retrieval-augmented generation, and computer vision. My work spans multi-agent applications, scalable ML pipelines, backend infrastructure, and polished frontend experiences, with production deployments on AWS. I have contributed to visual-inspection systems at InspectifAI and OCR tooling at DATEV, and hold an M.Sc. in Computational Engineering from FAU Erlangen-Nurnberg.
01M.Sc. Computational Engineering
02AWS ML Specialty + Associate
03LLM, RAG + Computer Vision
Technology stack
Tools I use to ship
production AI.
From model development to APIs, interfaces, databases, cloud infrastructure, and deployment.
01
LanguagePython02
LanguageTypeScript03
DataSQL04
Machine learningPyTorch05
Machine learningTensorFlow06
Machine learningKeras07
Machine learningscikit-learn08
Generative AIHugging Face09
Generative AILangGraph10
Computer visionOpenCV11
BackendFastAPI12
BackendDjango13
FrontendReact14
FrontendNext.js15
DatabasePostgreSQL16
DatabaseRedis17
CloudAWS18
MLOpsDocker19
MLOpsKubernetes20
MLOpsMLflow