Location: Lviv, Ukraine
Company Overview
Robots & Pencils is an applied AI engineering firm building the next frontier of business architecture. We design and ship AI co-workers that integrate into enterprise operations and deliver measurable results for our clients. We're all in on AWS, combining deep UX capability with senior engineering talent to get AI into production fast and keep it there.
We've earned the trust of leaders across Consumer Products and Retail, Education, Energy, Financial Services, Healthcare, and Manufacturing and more, and earned a reputation as the nimble alternative to traditional global systems integrators. Founded in 2009, with delivery centers in Canada, the United States, Eastern Europe, and Latin America, we are smaller, faster, and more senior by design. Our teams average 15+ years of experience. We move fast, sweat the details, and build things that actually ship.
Position Overview
We're looking for a Data Scientist to build the forecasting layer that will feed into AI systems already running in production for our clients. This is a hands-on modeling role at its core — classical time-series and ML work on real, messy business data — with a “bridge to GenAI” component: as your models mature past the pilot stage, you'll help package them as a served tool that an LLM agent can call. You'll work as a close collaborator with the AI Engineer(s) who own the surrounding agent/MCP infrastructure, not behind a ticket queue, and take ownership of turning ambiguous business questions into well-defined, validated models.
What You'll Do
- Build and validate demand-prediction models using classical forecasting/ML methods — ARIMA/SARIMAX, Prophet, and gradient boosting (LightGBM/CatBoost) on pooled tabular data
- Engineer time-series features: lags, rolling statistics, calendar/seasonal encoding, and exogenous regressors
- Pull, clean, and reason about data from messy ERP/SQL extracts (Azure SQL, QAD-sourced), and define target variables from ambiguous business data in collaboration with stakeholders
- Work through open questions in the current plan — including whether regional weather-to-customer mapping is a useful exogenous signal — and help pin down scope (seasonal calendar candidates, SKU scope, history depth)
- Document modeling decisions, assumptions, and known limitations clearly, for the client and for internal handoff
- Good to have: as a model matures past the pilot stage, wrap it as a FastAPI service in Docker and help expose it as an MCP tool an LLM agent can call
What You'll Bring
- 3+ years of professional experience in data science or applied ML, with hands-on delivery of forecasting/predictive models in a production or client setting
- Strong classical forecasting/ML background — ARIMA/SARIMAX, Prophet, gradient boosting (LightGBM/CatBoost) on pooled tabular data
- Feature engineering for time series — lags, rolling stats, calendar/seasonal encoding, exogenous regressors
- Python: pandas/numpy/statsmodels/scikit-learn/Prophet; comfortable working from messy ERP/SQL extracts (Azure SQL, QAD-sourced) and defining target variables from ambiguous business data
Nice to Have
- Enough FastAPI + Docker experience to wrap a trained model as a callable, containerized service — doesn't need to design the deployment architecture from scratch, but should ship one without hand-holding
- Familiarity with MCP (Model Context Protocol) or any agent-tool-calling pattern — enough to understand how a served forecast becomes one more tool an LLM agent calls, and to design “graceful degradation” (explicit fallback flags, never a silent/unearned number) for that consumer
- Basic Azure exposure (App Service/Container Apps, or just Azure SQL) — enough to get a demo host running, not full MLOps
- MLOps adjacents: model versioning, drift/retraining triggers — relevant once this POC graduates past pilot notebooks
- Weather/exogenous-data integration experience (regional weather-to-customer mapping) is a direct plus, since that's literally one of the open hypotheses in the current plan
- Consulting/client-services comfort — ambiguous scope, shifting priorities, stakeholder sign-off dependencies (seasonal calendar candidates, SKU scope, and history-depth confirmation are all still open questions in the plan)
You'll Do Well Here if You Are
- A doer. You see something broken and fix it. You'd rather move on clarity than wait for certainty.
- A fast learner who knows you don't know everything. The AI landscape changes weekly. You're senior enough to know better and curious enough to keep learning anyway.
- Direct in a way that makes the work better. You give honest feedback. You'd rather have the hard conversation than blow smoke.
- Obsessed with craft. You know genius is in the details. You ship exceptional, not perfect, and you don't put your name on work you wouldn't stand behind.
- Built for ownership. You honor commitments, admit mistakes fast, and back your teammates when a decision costs something. No handoffs, no finger-pointing.
- All in. You treat clients' businesses like your own. You take the work seriously without taking yourself seriously.
- Resourceful when the budget, timeline, or team is tight. Constraints don't slow you down. They sharpen you.
- Glad to be in the room with people who care as much as you do. Our teams average fifteen-plus years of experience. We hire people who push each other to do better work.
We Offer
- 20 days of paid vacation.
- 15 days of unpaid vacation.
- All official public holidays off.
- 5 paid sick days without a doctor's certificate.
- Medical insurance.
- Military draft deferment (reservation) support.
- Cooperation under a gig-contract.
Why Join Robots & Pencils?
At Robots & Pencils, we don't just build software — we solve meaningful business problems through creative technology. You'll work alongside passionate engineers, designers, data scientists, and strategists in a remote-friendly, collaborative environment. With our expansion into Ukraine, now is the perfect time to join our growing global team and contribute to high-impact work from day one.
