RAG
// Retrieval pipelines with measurable recall, grounding, and failure cases
[AI SYSTEMS]
[AGENTS / RAG / EVALS]
Emanuel-Ionut Otel
// i make llms do useful work
> I build LLM systems end to end: agents, RAG, evals, observability, and the backend code that keeps them alive. Give me a messy workflow and enough logs.
├─10k+ monthly chatbot sessions
├─5.8k PDFs turned into eval data
└─500x cheaper job classification
> No evals, no traces, no clue.
├─Cluj-Napoca / Remote
├─Python / FastAPI / Next.js
└─Agents / RAG / Evals / Traces
Teams + clients
I've built across robotics, aviation, voice AI, and enterprise software.
Research code. Backend systems. Internal tools. Customer-facing agents. Different constraints, same debugger.
Publicis GroupeSpiraxVorwerkEmersonE.ONForeFlightBillund AquacultureLifeline RoboticsContactLoopRebelDotBinarCodePublicis GroupeSpiraxVorwerkEmersonE.ONForeFlightBillund AquacultureLifeline RoboticsContactLoopRebelDotBinarCode
BinarCodeRebelDotContactLoopLifeline RoboticsBillund AquacultureForeFlightE.ONEmersonVorwerkSpiraxPublicis GroupeBinarCodeRebelDotContactLoopLifeline RoboticsBillund AquacultureForeFlightE.ONEmersonVorwerkSpiraxPublicis Groupe
[SKILL_MATRIX]
// The tools I reach for when an LLM demo needs to become software.
[CORE]
LLMs
// Model integration, structured outputs, fine-tuning, and evals
Agents
// Tool-using workflows with bounded loops, retries, and traces
LLMOps
// Prompts, traces, datasets, evals, and latency in one loop
Beyond_The_Model
CI/CD PipelinesUI PrototypingLogging & TracingDeployment AutomationAI System ReviewsTechnical ScopingFast Feedback LoopsTechnical CommunicationADRs
Core_Stack
// AI & Machine Learning
MCPsVector DBsPrompt Engineering
// Backend Engineering
PythonFastAPIPostgreSQL
// Frontend & Tooling
ReactNext.jsTypeScript
// DevOps & Infra
DockerKubernetes
// More_Tools
STT/TTS,Deep Learning,Reinforcement Learning,Git,Bash,Redis,Pydantic,Async APIs,WebSockets,Tailwind,Shadcn/UI,Playwright,AWS,Azure,Linux,Helm Charts
> 16 more modules loaded