Senior Data Scientist

A senior individual contributor owning the data science work end to end: deciding what to measure, getting the data, building the models, defending the conclusions. You work directly with traders, deal economists, finance and treasury, and the founders — no account manager in between.

We are hiring for a sector, not for one narrow problem. Assignments come from business stakeholders and they vary; what stays constant is the industry and the independence expected. The team can grow with one or two junior data scientists reporting to you.

There is a data engineer and an established data lake, but not every source is organised that way — a meaningful part of the work still starts in raw tables and unprocessed exports.

What the work looks like

The system we are building exists so that anyone involved in a deal can reconstruct the P&L between any two points of the chain, at any moment, on an agreed and versioned allocation basis. That is where the role sits: P&L, cost and financial transactions. Operational data is an input to margin, not a subject of its own.

This is operational deal P&L, reconciled to but not identical with the statutory accounts; the general ledger and month-close stay in the accounting system.

Scope

• P&L reconstruction and margin attribution across the deal lifecycle — netback at entry, provisional P&L, final P&L after final pricing, actual costs, demurrage and claims — with a defensible account of every delta, and the ability to reproduce the P&L as of a past date
• Forecasting where it pays on the financial side: provisional cost lines invoiced later, cost and margin per tonne, demurrage and claim exposure, FX and open pricing-period exposure, working capital and cashflow
• Anomaly and fraud detection on financial data: invoice lines inconsistent with contract terms, cost booked to the wrong deal or period, duplicate invoices, off-market pricing between related parties, cost reallocation that improves one deal at another's expense
• The reporting layer for finance and risk — margin, cost per tonne, exposure, receivables, cash position — with single agreed definitions where trading, economics and finance disagree
• Turning vague business problems into a research plan, and results into a recommendation someone can act on
• Incomplete and contradictory financial data as the normal case: provisional costs never trued up, invoices arriving months late, deals kept open for years by claims

Requirements

• 7+ years in data science, with senior ownership of your own workstreams
• Oil & gas industry experience — at least two years of delivered projects in trading, refining, logistics, storage or distribution. You can read a pricing clause, an invoice with its cost lines and a demurrage claim without a translator, and you know why provisional and final invoices for the same cargo legitimately differ. Which segment matters less. What matters is physical hydrocarbon volumes and index-linked pricing — power or utilities experience alone does not cover that
• Working with coding agents, seriously. Agents write most of the code here. Bad agent output in analysis fails silently: a join fans out rows, a dropna removes half a population, a future-dated column leaks into features. We hire for verification habits, not speed — row-count and population checks, reconciliation against a number someone already trusts, held-out evaluation, a traceable path from raw data to any figure a decision-maker sees
• Strong Python and SQL, and readiness to work in raw data yourself: unfamiliar schemas, undocumented tables, exports never meant for analysis
• Forecasting and time-series modelling on real business data
• Experience taking models into production and operating them; close to essential for us
• Numbers that carry weight. Your output feeds invoices, hedges and shareholder reporting: reconciliation to the accounting system, a versioned and documented method, and an explanation the person signing the invoice can follow. Accuracy that cannot be explained loses to accuracy that can
• English sufficient to present findings and defend methodology
• Professional proficiency in Russian

What we offer

• Ownership of the data science work, and room to shape how it is done
• Direct access to founders and decision-makers; short feedback loops
• Generous AI tooling budget — frontier models are a working expense, not a perk
• Competitive compensation, discussed individually