Home / Gaming

The same models, pointed at games.

Games produce some of the richest behaviour data anywhere — sessions, matches, purchases, progression and social graphs. Our recommendation and interest-evolution research maps directly onto player modelling, so studios can act on player behaviour instead of just logging it.

Where it applies

Six ways the recommendation stack shows up in games.

Player Interest & Behaviour Modelling

Model how each player's taste in genres, modes, maps and items evolves across sessions and titles over time.

Matchmaking & Ranking

Rank opponents, teammates, lobbies and content using graph and sequence signals — not a single static skill number.

Churn & Retention Prediction

Spot players drifting away before they leave, with confidence values that LiveOps and CRM teams can act on.

In-Game Store & Content Recommendation

Top-K recommendations for items, bundles, cosmetics, events and modes, tuned to each player's recent behaviour.

Player Segmentation & LiveOps

Cohort-aware personalisation for offers, events, difficulty, onboarding and re-engagement campaigns.

Cross-Platform Player Identity

Connect the same player across PC, console, mobile and storefronts, and model behaviour that spans devices.

Same stack Powered by the same GNN-Transformer hybrid architecture and H200 / B300-class GPU compute — the graph is players, items, sessions and platforms; the sequence is the player's recent actions.
Why games

A near-perfect match for graph + sequence modelling.

Player behaviour is relational (who plays with whom, which items pair with which modes) and sequential (what happened this session versus last month). That is exactly the structure our cross-platform interest models are designed for — which is why the research transfers with very little adaptation.

Graph

Social & item

Players, squads, items, modes and maps as connected structure.

Sequence

Session flow

Recent matches and actions signal current intent.

Tabular

Progression

Level, spend, tenure and account features.

Output

Actionable

Churn risk, recommendations and match quality.

For studios & publishers

Turn player behaviour data into decisions.

Matchmaking, retention, personalisation or store recommendation — we can scope a research plan around your data.

Start a conversation →