Data Scientist — AI & Decision Intelligence (domain lead)

CBS Systems has built commercial analytics for FMCG brand owners and retail chains for 20 years. SightPulse (www.sightpulse.ai) is the agentic decision intelligence platform that grew out of that work — it is live with international brand owners and leading regional retailers, and it is where most of our engineering now goes.

We are looking for one person to join ramping up data science side of it: the models in the product, the models in client projects, and eventually lead the team that builds them.

What you would do

• The data science domain. Methods, modelling standards, what is good enough to ship and what is not. You set the bar.
• Models inside the product. The predictive and statistical layer our decision agents reason over — demand and sales forecasting, segmentation, promotion effects, opportunity scoring.
• The data science component of client projects. Real commercial problems with the client's own data, turned into decisions people actually take.
• Production, not notebooks. Versioning, lineage, retraining, drift monitoring. A model nobody maintains is not finished.
• A team, over time. You start hands-on and hire and lead as the work grows. That is the intended path, not a vague promise.

What we expect

• 5+ years in data science or applied advanced analytics, with models that ran in production and that you kept running
• Domain experience in FMCG, CPG or retail — strong preference. Sales, digital & traditional Marketing, Market data. It shortens everything.
• Time series and forecasting at scale
• Strong Python — pandas/Polars, scikit-learn, statsmodels — and code others can read
• Strong SQL, including why a query is slow on a large table
• Comfortable owning the pipelines that feed your models (ETL/ELT, lakehouse layers, data quality)
• Able to explain a result to a commercial director, not only to a peer
• Judgement about when a simple baseline beats a complicated model
• Serbian and English

If you are close but not certain you tick every box, write anyway. We would rather have the conversation.

Useful, not required

• LLMs in production: analytical agents, RAG over client data, evaluation and guardrails
• Orchestration (Airflow, Dagster, dbt) and Azure, AWS or GCP
• Spark, DuckDB or similar
• Mentoring

Why this one is worth a look

• We do not do outsourcing. We own our products end to end, so your work stays yours and you see what it does in the market.
• Data at real scale and real messiness — multiple sources, multiple countries, years of history.
• Statistics first: every recommendation the platform makes is traceable back to the model and the data point behind it. That constraint is the interesting part of the job.

Perks & benefits

• Private health insurance from day one
• Training and conference budget
• Niche mentorship from people who are genuinely champions in their area, inside and around the company
• Baby bonus for new parents
• An office on Kosančićev venac with a view over the Sava — one of the best locations in Belgrade
• Fruit and snacks in the office, on us
• Organised teambuildings and events through the year

How to apply

Linked-in or write to tara.savic@cbs.rs. Skip the technology list — tell us in a few lines which models you put into production, what they were used for, and what happened to them afterwards. That tells us more than a CV.

Izvor: https://www.linkedin.com/jobs/view/4465565295/

Nivo: senior