I'm a full-stack AI engineer who spent years as a cardiac surgery ICU, pediatric ICU, and ER nurse before I ever wrote code. Now I build the AI systems clinicians actually need — because I've been the person on the other end of a bad prior auth call, a missed handoff, or a documentation system that got in the way of patient care.

• 🔭 Currently building autonomous, evaluation-first AI agents for prior authorization and clinical workflows
• 🌱 Deepening MCP, LangGraph orchestration, and agent evaluation (Langfuse, Evidently AI, LLM-as-judge)
• 💡 Interested in: healthcare AI, autonomous agents, agent evaluation & observability, FHIR interoperability
• 🎯 Goal: build AI systems clinicians can actually trust — not just accurate in a demo
• 📫 Reach me: senay@senayyakut.dev · LinkedIn

_{Auto-pulled from dev.to/senaykt}

Languages

Backend & APIs

Frontend

AI / Agents

Evaluation & Observability

Vector Databases

Infra & Databases

Healthcare Interoperability

A clinician-facing voice agent that automates prior authorization requests. Validated by a health-tech CEO who asked to scale it nationwide after seeing it work on real clinical scenarios.

xAI Grok Voice LangChain FastAPI Next.js FHIR R4

An autonomous prior-auth agent built to be trusted, not just accurate. Confidence scoring routes uncertain decisions to human review; retry logic and full audit trails make every autonomous decision traceable.

LangChain FastAPI Langfuse Evidently AI FHIR R4

A conversational database assistant with a hybrid two-tier memory system. Turns natural language into SQL, with Redis for short-term conversation memory and Mem0/Qdrant for long-term semantic memory, plus ChromaDB-backed RAG for schema retrieval.

FastAPI LangChain ChromaDB Qdrant Mem0 Redis

AI-powered clinical decision support using RAG + GPT-4 — automated SOAP note generation, differential diagnosis support, ICD-10 coding, and allergy/contraindication safety checks.

Python FastAPI OpenAI GPT-4 FAISS React RAG

RAG-powered research agent grounded in NIH, CDC, and WHO sources, with multi-pass agentic search and full source citation on every claim.

Python Contextual AI Claude RAG Pipeline

An 8-layer defensive security system for AI agents: device fingerprinting, behavioral monitoring, prompt injection detection, memory integrity checks, and audit logging.

Python Security Architecture Prompt Injection Defense

Sole founding engineer at OptiKube, a B2B Kubernetes optimization platform — owned architecture, infrastructure, and delivery end-to-end from zero, across the full stack.

• Speaker, PyLadies SF — "From ER Nurse to AI Engineer: Building Real Healthcare AI Systems with RAG & Agents"
• Published, Harvard Data Science Review
• Winner, SF Builders Collective Agent Hack Day