At Centiro we are constantly on a journey of discovery. We think outside the box and strive to stretch the boundaries of possibilities. We create something unique everyday – something which is more than a software company. We believe in innovation and the strong power of collaboration. Our ambition is to deliver the best solutions for our customers and at the same time be the best possible place to work. Do you want to be a part of our continued journey?

Background

In last-mile logistics, carriers report shipment progress through event updates such as “Picked up,” “In transit,” or “Delivered”. These updates, known as carrier events, vary across carriers in naming, frequency, and sequence. This inconsistency makes it difficult to compare or analyze shipments systematically. The project aims to design an algorithm for cross-carrier event sequence alignment and canonicalization, mapping heterogeneous tracking events into a unified shipment lifecycle. The work combines three challenges: 

1. Building a weakly-supervised labeler to map raw event names into standardized phases, 

2. Learning an alignment method that can handle directional graphs (with cycles) and reordered events,

3. Inferring a set of partial order of events that explains most observed sequences while tolerating noise. 

Deliverables include algorithms, a small annotated dataset for evaluation, and an analysis of alignment accuracy, order fidelity, and downstream impact on prediction tasks. The project balances applied value (cross-carrier analytics) with a clear theoretical challenge (sequence alignment under noise and partial orders).

Once events are canonicalized and aligned into shared lifecycles, the learned partial-order models give you a picture of the normal flow across carriers, a profile for the carrier. This makes it possible to automatically flag deviant behaviors: sequences that contain unexpected transitions (e.g., Delivered before Pickup, Returned after Pickup), skip critical phases, or insert unusual detours not observed in the majority. Such deviations can indicate errors in carrier data, integration issues, or disruptive real-world processes (lost parcels, failed pickups, misrouted shipments). By quantifying how far a given sequence deviates from the canonical order—using edit distance, violation counts, or anomaly scores—the system can surface problematic shipments in real time for investigation.

Required skills

• Solid foundation in mathematics, particularly probability theory and optimization

• Machine learning (supervised and unsupervised methods)

• Algorithms and data structures

• Strong proficiency in Python for data processing and algorithm development

Preferred skills

• Graph theory and sequence alignment concepts

• Experience with cloud platforms (e.g., GCP) for data storage and model deployment

• SQL for data querying and manipulation

When
Spring of 2026

You are…
A student studying Data Science and AI, Mathematics, Computer Science or similar.