Requirements:

Strong Python (typing, tests, packaging) and web/service basics (FastAPI, Docker).
Hands-on building LLM apps with one or more: LangChain, LangGraph, AutoGen, Dify (or equivalents). RAG end-to-end experience: corpus prep embeddings retrieval re-rank context assembly* measurable gains.
Evaluation mindset: design offline eval sets and release gates; reason about faithfulness vs relevancy; control p95 latency & $/session.
Cloud proficiency in AWS or Azure (using at least one in production).
Comfort with observability (OpenTelemetry/LangSmith/Prometheus/Grafana or similar).
English B2+; clear written & verbal client communication.

Responsibilities:

Design & build LLM applications: RAG over client KBs, multi-tool/agent graphs, prompt strategies, and voice UX (STT/TTS).
Implement retrieval & data flows: ingestion, chunking, metadata/embeddings, hybrid retrieval, re-ranking, caching.
Own evaluation & quality: latency/cost budgets, A/Bs.
Automate with n8n (or similar): connect services, data, and back-office workflows; expose reliable admin ops for non-engineers.
Ship pipelines: Airflow for orchestration; dbt for transformations/semantic layer where relevant.
Productionize: tracing/observability, feature flags, safe rollouts, secrets & config hygiene.
Collaborate with clients: discover scope, explain trade-offs, demo progress; optional presales spikes/estimates.

Vector DBs: pgvector, Qdrant, Pinecone, Azure Cognitive Search (vector); rerankers (BGE/monoT5/Cohere/ColBERT).
Voice: Azure Speech, Whisper; VAD/barge-in, turn-taking logic.
Data stack: dbt, Airflow, event streams; data contracts.
Workflow: n8n design for non-tech ops; governance of low-code automations.
Frontend for AI UX (React/Next.js, TypeScript).
Compliance awareness for EdTech (FERPA/COPPA) and Healthcare (HIPAA/GDPR).

Olesia T.

ML/AI Work links you to the employer's original posting — always verify the details there before applying.