DrepoTech's role
DrepoTech shaped the product strategy, operational dashboard, agentic decision loop, explainability model, and direction from city prototype toward a configurable municipal platform.
Civic technology & AI
From one city prototype to an agentic operating system for municipal waste
A flagship smart-city platform concept that helps municipal teams predict overflows, coordinate vehicles, adapt routes to live conditions, and understand every AI-assisted recommendation.
Civic technology & AI
GigaBin
The challenge
Municipal waste collection is not a static routing exercise. Fill levels, contamination alerts, traffic, truck capacity, geography, and service priorities change throughout the day. Operators need one dependable way to see what requires attention, understand what the system recommends, and remain in control of the final decision.
DrepoTech's role
DrepoTech shaped the product strategy, operational dashboard, agentic decision loop, explainability model, and direction from city prototype toward a configurable municipal platform.
Status context
The current evidence is a flagship prototype developed around Cluj-Napoca; wider municipal deployment is a product direction, not a claimed live rollout.
Project objectives
The business priorities and customer needs that shaped the delivery.
Strategy
Professional delivery starts by making the important product, content, experience, and ownership decisions visible.
Design around the professional decision or workflow rather than presenting model output as the product itself.
Give the user context, review points, clear actions, and an understandable boundary between assistance and authority.
Account for uncertainty, revision, failure, and escalation instead of assuming every generated result is usable.
Explain recommendations, preserve human oversight, and expose the information needed to judge the system.
The solution
GigaBin brings observation, overflow prediction, urgency scoring, traffic-aware routing, vehicle capacity, truck assignment, and operator controls into one coherent decision loop: observe, predict, decide, act, explain, and learn.
Specialized AI roles contribute to each recommendation while the interface exposes the relevant evidence and rationale. The initial prototype was developed around Cluj-Napoca during the Erasmus+ BIP 2026 at Babeș-Bolyai University, with academic guidance from the international programme team.
The next product direction expands that validated workflow into a configurable platform for municipalities and waste operators that need cleaner streets, more responsive collection planning, and accountable human oversight.
Delivered scope
Technology and platform
Delivery approach
The exact activities change by engagement, but the delivery sequence protects clarity, quality, and ownership.
Specify the user, input, desired output, decision boundary, failure cost, and success criteria.
Test the interaction with representative scenarios before expanding model or platform complexity.
Connect model behavior to the product interface, data, controls, feedback, and operational workflow.
Review output quality, edge cases, human control, performance, cost, and the path for continued improvement.
Delivered value
How the delivered scope supports the business, its customers, and the people using the system.
Observation, prediction, urgency, routing, assignment, action, and explanation are represented as one connected workflow.
The interface exposes relevant evidence and rationale so operators can judge rather than blindly accept a recommendation.
The city-specific work now provides a foundation for evaluating configuration, integration, and governance needs across municipalities.
More work
Marketing website
Njeriu i Sharrit
A focused digital home for a public personal brand
B2B e-commerce engineering
Proteini.mk
End-to-end e-commerce engineering delivered through a B2B partnership
E-commerce & 3D
Arcavo
Furniture e-commerce with interactive 3D and in-room previews
Start a conversation
Share the operating problem, current constraints, and what a useful result should change. We’ll recommend the most responsible next step.