Senior AI Engineer

The role
As a Senior AI Engineer, you'll own the design and delivery of the AI systems at the core of Epistak's
product. You'll take LLM-powered features from prototype to reliable production service, set the bar for how we evaluate model quality, and help shape our engineering culture and early hiring. This is a hands-on individual contributor role reporting to the Founding Architect and the Founder.
What you'll do
Design, build and ship production AI features using large language models, retrieval-augmented generation (RAG) and agentic workflows.
Build agents that connect to enterprise IT and security systems, unify their data into shared context, and reason across it for root-cause analysis, patch prioritization and alert triage
Make every conclusion explainable, with the evidence behind it, and design automated actions that stay within defined policies and approvals
Own the architecture of our AI stack: model selection, orchestration, retrieval, caching and guardrails
Build evaluation pipelines (offline test sets, LLM-as-judge, human review loops) and use them to drive decisions
Monitor models in production for quality, latency, cost and drift, and fix issues before customers notice
Fine-tune or adapt models where off-the-shelf APIs fall short
Work with product and design to turn ambiguous problems into well-scoped AI solutions
Review code and design docs, and set the technical bar for the engineers who join after you
Keep up with the field, spot emerging patterns and techniques, and bring them into the product when they make a real difference
What you bring
The ideal candidate started in data engineering (analytics and data modelling), moved into ML
engineering, and has built LLM-based workflows since GPT arrived. That path gives you a feel for the
data, the models and the workflows together.
A data engineering foundation: strong SQL, data modelling, and building pipelines on warehouses or lakes (e.g. Snowflake, BigQuery, Databricks)
Hands-on experience shipping LLM-based workflows: prompt design, RAG, tool use, agents and multi-step pipelines
Strong machine learning fundamentals: supervised learning, embeddings, model evaluation metrics, overfitting and bias, and when a classical model beats an LLM
Strong Python, plus experience with PyTorch, Hugging Face, scikit-learn or similar
Experience with vector databases and search (e.g. pgvector, Pinecone, Weaviate, Elasticsearch)
A rigorous approach to evaluation — you measure quality rather than eyeballing it
Solid backend fundamentals: APIs, data pipelines, testing, CI/CD
Experience deploying on a major cloud (AWS, GCP or Azure) with Docker and Kubernetes
Nice to have
Startup experience
Experience with knowledge graphs or other ways of modelling relationships between entities
Experience with IT operations or security data (ITSM, SIEM, CMDB, vulnerability scanners)
Experience fine-tuning models (LoRA, RLHF/DPO) or training smaller task-specific models
Background in MLOps tooling: experiment tracking, model registries, feature stores
Experience optimizing inference (quantization, batching, serving with vLLM or similar)
Publications, open-source contributions or a portfolio of AI projects

Work arrangement
Hybrid