Helsinki, Uusimaa, Finland · On-site
Teach the engine to reason, then ship it At Fluxion, modeling and AI are the same problem. You'll work on the core engine that evaluates, simulates, and reasons over complex graph-based business models. You'll also build…
Skills: Python, Data Science, Artificial Intelligence, Mathematical Modeling, Statistics
Sign up with Clera and we'll reach out the moment a role actually fits you — no more spraying applications into the void.
You will design and implement the mathematical core of the modeling engine and build AI agents that interpret intent and explore scenarios. This role requires taking ideas from mathematical concepts to production-ready code within the main product codebase.
Candidates should have a strong quantitative background in fields like mathematics, physics, or statistics with proven experience building production AI systems. You must be proficient in Python and have deep knowledge of probability, optimization, and LLM-based agent orchestration.
At Fluxion, modeling and AI are the same problem.
You'll work on the core engine that evaluates, simulates, and reasons over complex graph-based business models. You'll also build the AI agents that use it to explore assumptions, propose scenarios, and explain what they find.
This is an end-to-end role. You will not hand a notebook to an engineer. You own the path from mathematical idea to production code, in the same codebase as everyone else.
This role sits at the intersection of:
Applied mathematics and statistics
Semantic and causal modeling
LLM-based agents and reasoning systems
Simulation and scenario analysis
Production engineering
We are building infrastructure for reasoning over structured models, and the agents that reason with it.
Design and implement the mathematical and statistical core of our modeling engine, covering forecasting, uncertainty, sensitivity, and causal propagation
Turn modeling ideas into production code in the product codebase, not into prototypes someone else rewrites
Build AI agents that operate on top of the engine: interpreting intent, constructing and modifying models, and running scenario exploration
Design the interface between deterministic computation and probabilistic reasoning, deciding what the engine must guarantee and what the agent is allowed to infer
Develop evaluation frameworks for agent behaviour, covering correctness, reliability, regression, and failure analysis
Make the output explainable, with traceable assumptions, visible causality, and defensible numbers
Work directly with users' real modeling problems and translate them into engine capabilities
Come from a quantitative background (data science, mathematics, statistics, physics, operations research, or similar) and have since moved into building AI systems
Are a genuinely end-to-end data scientist: you ship, review, test, and maintain code that runs in production
Have strong foundations in probability, statistics, optimisation, and numerical methods, and know when a model is wrong rather than merely unconverged
Have built with LLMs beyond prompting, including tool use, structured output, agent orchestration, retrieval, and evaluation
Think rigorously about correctness and determinism, and are uncomfortable shipping a system whose failure modes you cannot describe
Can reason about the semantics of a problem and not just the metrics, about why a relationship holds and not only that it correlates
Communicate clearly about uncertainty with people who need to make decisions under it
Python
Polars, NumPy, SciPy
PostgreSQL, SQLAlchemy
FastAPI
Additionally, experience with any modeling packages, such as scikit-learn, Prophet, XGBoost etc. and any AI agent orchestration and evaluation frameworks is beneficial and potentially applicable.
Background in Computer Science, Mathematics, Physics, Statistics, or a similar quantitative field
Experience with financial modelling, forecasting, optimisation, planning, or simulation systems
Experience designing evaluation harnesses for LLM or agent systems
Experience with causal inference, Bayesian methods, or probabilistic programming
Experience with graph-based representations, constraint systems, or symbolic computation
Have built a product feature end to end, not only analysis or research output
Have shipped work where you both derived the method and wrote the production code behind it
Prefer designing the system underneath the pipeline to tuning the pipeline that already exists
Are at home in problems that are not yet well specified, on a stack that is still changing
Fluxion is a decision intelligence platform that empowers forward-looking teams to answer their toughest "what-if" questions with confidence. Using AI-native scenario modeling, Fluxion lets organizations explore assumptions and simulate outcomes across financial and operational models in real time — without rebuilding spreadsheets or rigid tools. Its semantic approach to decision modeling keeps assumptions explicit, preserves causality, and delivers traceable, explainable insights so teams make better decisions faster. Designed for high-impact decisions under uncertainty, Fluxion replaces fragile, manual exploration with a dynamic environment where curiosity is cheap, safe, and continuous.
IS UW IDEE KLAAR VOOR DE START? Ideeën zijn de innoverende kracht van uw bedrijf. Maar welk idee is nou echt kansrijk bij uw klanten? Eentje die ze met smacht omarmen? Fluxion biedt op basis van Business Model Generation een passende aanpak voor uw idee. In 1 dag scheiden we samen met u én uw klanten het kaf van het koren. In 2 dagen weet u ook of uw idee commercieel en organisatorisch haalbaar is. En in 3 dagen bent u met een concrete aanpak klaar voor de start. Voor de vooraf afgesproken prijs. Wij houden van klanten. Sinds 2000 maakt Fluxion organisaties vanzelfsprekend klantgedreven. We vertalen uw innovatieve ideeën in waardevolle concepten. Dit doen we in co-creatie met u en uw klanten. Dankzij onze heldere en doortastende aanpak weet u snel of u waar kunt maken wat u aan klanten wilt beloven. Zodat uw product of dienst straks ook echt iets oplevert. Voor uw klant en voor u.
Offices: Bruggestraat 49-1, Harderwijk, 3841 CM, NL
IS UW IDEE KLAAR VOOR DE START? Ideeën zijn de innoverende kracht van uw bedrijf. Maar welk idee is nou echt kansrijk bij uw klanten? Eentje die ze met smacht omarmen? Fluxion biedt op basis van Business Model Generation een passende aanpak voor uw idee. In 1 dag scheiden we samen met u én uw klanten het kaf van het koren. In 2 dagen weet u ook of uw idee commercieel en organisatorisch haalbaar is. En in 3 dagen bent u met een concrete aanpak klaar voor de start. Voor de vooraf afgesproken prijs. Wij houden van klanten. Sinds 2000 maakt Fluxion organisaties vanzelfsprekend klantgedreven. We vertalen uw innovatieve ideeën in waardevolle concepten. Dit doen we in co-creatie met u en uw klanten. Dankzij onze heldere en doortastende aanpak weet u snel of u waar kunt maken wat u aan klanten wilt beloven. Zodat uw product of dienst straks ook echt iets oplevert. Voor uw klant en voor u.
Offices: Bruggestraat 49-1, Harderwijk, 3841 CM, NL
