Accelerates AI model training by ~35% with Vertex AI
Lowers compute costs by ~25% with Compute Engine
Enables generative AI trade agents with Vertex AI
Unifies hybrid cloud visibility with Cloud Logging
DiMuto uses AI agents to cut training time by 35%, costs by 25%, and reduce global food waste.
DiMuto uses AI agents to cut training time by 35%, costs by 25%, and reduce global food waste.
Digitizing the global agri-food trade
Our setup was reliable but increasingly fragmented. To move faster, we needed to unify model training, orchestration, and monitoring. We wanted to bring everything into one ecosystem so we could focus on refining our AI rather than managing infrastructure.
Thet Hlaing
Technical Lead, DiMuto
20-30% of the world's food production is lost or wasted. For DiMuto, solving this requires opening up the supply chain's 'black box' — the critical period between a packhouse and a retailer where visibility often vanishes. DiMuto tracks individual produce items, digitizing every step to ensure quality and resolve disputes.
Offering this holistic overview within a fragmented, hybrid cloud environment was a challenge. Critical data and machine learning workloads were siloed, creating friction for engineering teams. Training the AI models essential for spotting fruit defects was slow, and maintaining reliability across disparate systems left the team juggling operational maintenance instead of advancing the platform.
To evolve from tracking goods to actively predicting their quality, DiMuto needed a unified, scalable foundation.
Automating quality control with Vertex AI
DiMuto expanded its Google Cloud footprint to harness its generative AI tools while maintaining full compatibility with critical open-source frameworks such as FastAPI and ChromaDB.
With heavy compute and machine learning workloads now running in this unified environment, the team relies on Cloud Logging to provide complete visibility across its hybrid operations.
Now, when a carton of avocado arrives at a packhouse, it is scanned and instantly analyzed by Vertex AI. The system detects skin defects and grades quality in real-time, then hashes the results onto a public blockchain hosted on Compute Engine for immutable traceability.
User expectations have changed. It's no longer enough to show search results; people want the answer immediately. We are using Gemini models to let users cut through the noise of complex trade documentation and get straight to the information they need.
Jordon Ho
Head of Tech Ops, DiMuto
The platform can also "speak." Using the Google Agent Development Kit (ADK) and Vertex AI Agent Builder, DiMuto built a Trade Contract Agent supported by Gemini models. Instead of hunting for files, a user simply asks, "Where is my bill of lading?" Behind the scenes, each trade contract is converted into a structured markdown document and registered with its S3 path.
The ecosystem delivers performance and security:
Building a waste-free global supply chain
Our collaboration with Google Cloud is transformative. It enables us to reinvent our solutions, particularly in AI development. The collaboration's spirit and the support we receive help us build the foundation for a waste-free global supply chain.
Thet Hlaing
Technical Lead, DiMuto
Centralizing complex workloads on Google Cloud yielded immediate efficiency dividends. With Vertex AI, the team reduced by roughly 35% compared to our previous pipeline, allowing data scientists to iterate on produce quality models (such as Papaya Quality AI) at a faster rate, and freeing up time to refine the intelligent agents that power the platform.
By optimizing our training pipeline with Preemptible VMs on Compute Engine for suitable workloads and redesigning parts of the pipeline to be interrupt-tolerant, DiMuto achieved roughly 25% cost savings compared to comparable GPU workloads in its previous environment. These savings were reinvested in R&D to accelerate new innovations such as the Trade Contract Agent.
Looking ahead, DiMuto plans to deepen its use of Vertex AI Agent Builder to create a fully agentic supply chain and support the company's ultimate vision: a transparent global trade ecosystem that minimizes disputes and the environmental impact of food waste.
DiMuto digitizes the global agri-food supply chain using AI and blockchain to provide end-to-end visibility that resolves trade disputes and reduces food waste.
Industry: Agriculture
Location: Singapore
Products: Vertex AI, Compute Engine, Cloud Logging, Cloud Monitoring, Cloud Run, Cloud SQL, Cloud IAM, Artifact Registry
