Artificial Intelligence Engineer

We’re looking for an applied AI Engineer who can take models from notebook to production. You’ll design, build, and ship solutions around LLMs, RAG pipelines, and classic ML, working across data processing, modeling, evaluation, and deployment. If you’ve actually put AI features in users’ hands—and supported them in the wild—you’ll feel at home here.

What you’ll do

• Own end-to-end delivery of AI features: problem framing, data prep, modeling, evaluation, deployment, and monitoring.

• Build RAG systems using LangChain/LlamaIndex, embeddings, and vector databases (FAISS/pgvector/Milvus/Pinecone).

• Fine-tune or instruct-tune LLMs; perform prompt engineering, safety/guardrail design, and latency/throughput optimization.

• Implement inference services (FastAPI/Flask), batch and streaming pipelines, and robust evaluation harnesses (A/B, human-in-the-loop).

• Train and serve models using PyTorch or TensorFlow; convert/optimize with ONNX/TensorRT when needed.

• Instrument systems for observability (metrics, tracing, drift/quality monitoring) and write clean, tested, version-controlled code.

• Collaborate with product/engineering to scope milestones, de-risk unknowns, and document decisions.

Must-have qualifications (aim for ~70% match)

• 2+ years of hands-on AI/ML implementation (not just coursework); at least one model/LLM feature shipped to production.

• Strong Python; deep comfort with PyTorch or TensorFlow (preferably PyTorch).

• Practical experience with LLMs (open-source or hosted), embeddings, RAG, LangChain/LlamaIndex.

• Experience with vector databases (FAISS required; plus one of pgvector/Milvus/Pinecone).

• Building and exposing ML/LLM services via APIs (FastAPI/Flask) and containerizing with Docker.

• Data wrangling (Pandas/Polars), feature engineering, and model evaluation (including offline/online metrics).

• Solid software engineering practices: Git, code reviews, testing, linting, reproducible environments.

• Clear written and verbal communication; ability to translate requirements into technical execution.

Nice to have (great, not mandatory)

• Prompt-tooling and eval frameworks (Ragas, DeepEval, TruLens, or custom eval suites).

• MLOps: CI/CD for models, model registries, artifact/version management, feature stores.

• Performance optimization on GPU/CPU; quantization (bitsandbytes, GGUF), caching, batching, concurrency.

• Cloud experience (AWS/Azure/GCP), serverless (Lambda/Functions), and infra-as-code (Terraform).

• Search/retrieval: Elastic/OpenSearch, hybrid search (sparse+dense), chunking and indexing strategies.

• Security & privacy for AI systems; experience with on-prem/offline deployments.

• Classic ML (tree-based models, sklearn), NLP basics (tokenization, NER, summarization), and CV exposure (OpenCV).

• Databases and messaging: PostgreSQL/SQL Server, Redis, Kafka/RabbitMQ.

Tools you might use here

Python, PyTorch, TensorFlow, LangChain/LlamaIndex, FAISS/pgvector/Milvus/Pinecone, Azure OpenAI/OpenAI/Ollama, FastAPI, Docker, GitHub Actions, ONNX/TensorRT, Power BI, Pandas/Polars, PostgreSQL/SQL Server, Redis.

What success looks like (first 90 days)

• Ship a small LLM-backed feature end-to-end (design → deploy → measure).

• Stand up a RAG pipeline with proper evaluations and retrieval quality baselines.

• Improve latency/cost/reliability of at least one existing AI service via optimization or caching.

• Leave behind crisp documentation and dashboards so others can operate what you build.

Equal opportunity

We value skill, curiosity, and ownership. If you meet most of the “must-haves” and are eager to learn the rest, we’d love to hear from you.