Senior AI Engineer

Altimetrik Poland is a digital enablement company. We deliver bite-size outcomes to enterprises and start-ups from all industries in an agile way to help them scale and accelerate their businesses. We are unique in Poland's IT market. Our differentiators are an innovation-first approach, a strong focus on core development, and an ability to attack the challenging and complex problems of the biggest companies in the world.

La experiencia que se espera de los solicitantes, así como las habilidades y cualificaciones adicionales necesarias para este trabajo, se enumeran a continuación.

We are looking for a Senior AI Engineer to sit at the intersection of AI platform capabilities and Finance business users — translating ambiguous business problems into working AI solutions, deploying them into the customer's environment, and iterating in tight loops until they deliver measurable value.

What You'll Do

• Own end-to-end delivery of AI solutions for specific Finance business problems — from problem definition with the business stakeholder through deployment, adoption, and measurable outcome
• Translate ambiguous, evolving business requirements into working software — often without a formal spec, working directly with the Finance user who owns the problem
• Build vertical AI applications rapidly on top of the existing AI platform — leveraging Snowflake Cortex, Databricks Genie, RAG pipelines, and agentic workflows to solve narrow, high-value business problems
• Prototype in days, not months — get a functional Streamlit app, Databricks App, or lightweight web UI in front of business users within the first 1-2 weeks of engagement, then iterate based on real user feedback
• Deploy solutions inside the customer's regulated environment — respecting existing governance, RBAC, data residency, audit, and compliance constraints
• Instrument for measurement — every deployed solution ships with usage metrics, adoption tracking, and business outcome telemetry from day one
• Handle the "last mile" that makes AI actually usable — data quirks, business rule exceptions, edge cases, user training, change management, and adoption support
• Work directly with Finance business stakeholders — CFO office, FP&A, controllership, treasury, procurement — translating their language into technical solutions and back
• Build reusable patterns — after solving a specific problem, extract the reusable pieces (prompts, retrieval patterns, UI components, evaluation harnesses) into shared assets other FDEs can leverage
• Own the outcome, not just the code — if the business user isn't getting value, the job isn't done regardless of whether the code is deployed

Must-Have Technical Skills

• Full-stack AI application development — you can build a working end-to-end system with a UI, backend, and AI/LLM integration in weeks, not months
• Python + FastAPI/Flask for backend services, Streamlit / Databricks Apps / lightweight React for user-facing interfaces
• Direct hands-on with at least one of: Snowflake Cortex (Analyst/Search/Agents/LLM Functions) OR Databricks Genie (Genie Spaces, semantic models) — you know these products well enough to configure, tune, and integrate them into vertical solutions
• RAG pipeline construction — chunking, embeddings, vector search (Pinecone, pgvector, Chroma, FAISS, Azure AI Search, Snowflake Cortex Search), retrieval evaluation, grounding, citation
• LLM application frameworks — LangChain, LangGraph, LlamaIndex, or equivalent — with production usage, not tutorials
• Prompt engineering with evaluation discipline — you know how to design prompts, evaluate them against ground truth, iterate based on hallucination and accuracy metrics
• Cloud data platform fluency — Snowflake and/or Databricks at working depth, plus at least one cloud provider (Azure preferred given the Novartis environment, AWS/GCP acceptable)
• SQL and data modeling — enough to work directly with governed datasets and semantic models
• Git, CI/CD, and modern development workflows — you own the deployment path, not just the local development
• API design and integration — you can integrate your solution into existing enterprise systems (SAP, Workday, Coupa, ERP, etc.) via REST APIs
• Basic MLOps awareness — you understand model versioning, prompt versioning, evaluation harnesses, observability (LangSmith, Datadog, Application Insights) — enough to hand off your solution to the MLOps team cleanly

Must-Have Non-Technical Skills

• Direct customer/stakeholder communication — you can sit in a room with a CFO office user, understand what's frustrating them, and translate that into a technical roadmap without needing a business analyst intermediary
• Ambiguity tolerance — you're comfortable starting work with a vague problem statement and refining it through prototypes rather