Work we are proud of
Examples of our work across industries.
Demand forecasting engine
Challenge: Inventory levels relied on guesswork and seasonal peaks caused surprises.
Solution: We built a time-series model forecasting demand per product.
Support agent for customer service
Challenge: The support team was drowning in repetitive routine tickets.
Solution: We built an agent that resolves routine cases and escalates the rest.
Vision-based quality inspection
Challenge: Manual inspection was slow and defects slipped through.
Solution: We trained a model to spot defects on the line in real time.
Route optimisation
Challenge: Delivery routes were planned by hand and were inefficient.
Solution: We built an optimisation model accounting for time and capacity.
Automated document reading
Challenge: Processing invoices and contracts consumed enormous time.
Solution: We built an OCR and extraction pipeline for structured data.
Knowledge search for care guidelines
Challenge: Staff could not quickly find guidance in extensive documentation.
Solution: We built a RAG search that answers with citations.
Content classification and tagging
Challenge: A large content archive was unsorted and hard to search.
Solution: We built automatic classification and tagging.
Predictive maintenance model
Challenge: Equipment failures could not be anticipated and downtime was costly.
Solution: We built a model anticipating failures from sensor data.
Multilingual learning assistant
Challenge: Students needed support in several languages around the clock.
Solution: We built a conversational assistant grounded in course material.
Price estimation model
Challenge: Price estimates varied by appraiser and were inconsistent.
Solution: We built a model estimating price from property features.
Personalised recommendations
Challenge: Customers received generic offers that missed the mark.
Solution: We built a recommendation engine from behavioural data.
Real-time operations dashboard
Challenge: Production status was scattered across different systems.
Solution: We combined sources and built a real-time view.