Every liter of used cooking oil poured down a drain clogs pipes, pollutes waterways and wastes a valuable resource. Noovoleum saw an opportunity: pay people for that oil and turn it into feedstock for cleaner fuels. To make that model work at scale, Noovoleum partnered with CASA5 to build a modern cloud backbone on AWS, from smart collection boxes in the field to a multilingual AI assistant in the app.

About Noovoleum: used cooking oil recycling that pays

Noovoleum is a climate-tech company, with offices in Bandung and Singapore, that makes used cooking oil recycling as simple as dropping off a bottle. Through its UCOllect program, households, students and small food businesses bring their used cooking oil (UCO) to an automated UCOllect box, deposit it, and receive instant credit in the UCOllect app. Their tagline says it plainly: “Easy Oil Drop, Fast Cash.”

The collected oil does not go to waste. It becomes feedstock for renewable fuels, including sustainable aviation fuel (SAF). Energy companies across the region are taking notice. For example, Thailand’s Bangchak announced a UCollect Box pilot in Bangkok to supply its “Fry to Fly” SAF project.

The model creates value on two sides at once:

• A cleaner environment: waste oil stays out of drains, soil and rivers, and it replaces fossil feedstock in fuel production.
• New income: ordinary people and small businesses earn money from something they used to throw away, and the renewable-fuel industry gets a traceable, reliable supply.

The challenge: scaling a network of remote collection points

A recycling network is only as good as its operations. As Noovoleum added more UCOllect boxes in more locations, running them by hand became the bottleneck. The team faced:

• No real-time view of oil levels and oil quality at each box
• Manual, time-consuming data collection and reporting
• Slow response to maintenance needs and equipment issues
• Inconsistent data across locations
• High costs from unnecessary site visits and inefficient pickup routes
• A growing volume of customer questions, in more than one language, as the program expanded across countries

To grow, Noovoleum needed one central platform that could collect, process and analyze data from every box automatically, and a support experience that could keep up with its users.

Phase 1: A modern IoT stack for smart UCOllect boxes

CASA5 designed and delivered a secure, scalable IoT architecture on AWS in about three and a half weeks. Each UCOllect box became a connected edge device that reports its status to the cloud in real time.

At the edge

• Raspberry Pi 5 runs as the edge computer inside each box, hosting the AWS IoT Greengrass core software.
• A Uniboard interface connects the Raspberry Pi to the box’s sensors, which continuously measure operating data such as oil level and quality.
• AWS IoT Greengrass processes and caches data locally, so a box keeps working even when the internet connection drops, and less data needs to travel to the cloud.

In the cloud

• AWS IoT Core provides secure, scalable messaging and device management for every box.
• Amazon MSK (Managed Streaming for Apache Kafka) handles reliable, real-time data streaming.
• Amazon ECS on AWS Fargate, with images stored in Amazon ECR, runs containerized workloads without servers to manage, including Telegraf for data collection.
• Amazon Timestream for InfluxDB stores high-volume time-series sensor data.
• Amazon Managed Grafana turns that data into live dashboards and alerts for the operations team.

What changed: Noovoleum can now manage and configure edge devices remotely, monitor the status and performance of every box in real time, and push new software to boxes in the field without sending a technician. Pickups can be planned around actual fill levels instead of guesswork, and opening a new location becomes a repeatable, low-effort process.

Phase 2: A multilingual AI assistant with Amazon Bedrock

With more users in more countries, support had to scale too. CASA5 built a generative AI customer service assistant for the UCOllect app in two weeks, as a working proof of concept and a ready foundation for production.

The assistant answers questions about points, deposits, transaction history and box locations. It grounds every answer in Noovoleum’s own standard operating procedures (SOPs) and flags complex issues for escalation to the support team.

How it works

1. A user sends a message in the chat. Amazon API Gateway receives it securely.
2. An AWS Lambda function detects the user’s intent and language, and decides whether account data is needed.
3. Smart model routing keeps costs low: simple questions go to the fast, efficient Claude Haiku model, while complex questions go to Claude Sonnet. Routing prompts live in Amazon Bedrock Prompt Management, not hard-coded logic.
4. Amazon Bedrock Knowledge Bases, with Amazon Titan Text Embeddings and an S3 vector store, retrieves the most relevant SOP content (retrieval-augmented generation, or RAG), so answers are based on Noovoleum’s real policies rather than model guesswork.
5. The response streams back to the user in their preferred language, and Amazon CloudWatch captures logs and metrics.

The application layer runs in the AWS Jakarta Region (ap-southeast-3), close to Indonesian users, while AI processing runs in Sydney (ap-southeast-2), where the needed Bedrock features were available. This balances speed, cost and service availability.

What changed: Users get instant, accurate answers around the clock in their own language. The support team spends less time on repetitive questions and more on the cases that need a human. As Noovoleum enters new markets, it can add a language or update an SOP without rebuilding anything.

The business impact: a cleaner planet and new income

Together, the two projects give Noovoleum the technology foundation to grow its circular economy model with confidence:

• More oil recovered, less pollution. Reliable, well-maintained boxes mean more used cooking oil is captured instead of being dumped, and more renewable-fuel feedstock reaches producers.
• Trusted, traceable supply. Real-time sensor data on volume and quality helps give fuel producers the traceable UCO supply that SAF and biodiesel programs require.
• New income for communities. A smoother deposit-and-get-paid experience, with fast support, encourages more households and small businesses to earn from their waste oil.
• Lower operating costs. Remote monitoring and remote updates cut unnecessary site visits, and intelligent AI model routing keeps support costs efficient.
• Ready to scale across borders. A repeatable IoT rollout plus multilingual support supports Noovoleum’s expansion from Indonesia into markets such as Singapore and Thailand.

Why CASA5

CASA5 is an AWS partner based in Jakarta that helps Indonesian companies design, build and deploy solutions on AWS, from IoT and data platforms to generative AI. For Noovoleum, we delivered each phase in weeks, not months, using managed AWS services to keep operational overhead low and leaving a documented, extendable foundation the team can build on.

Frequently asked questions

What is used cooking oil recycling?

Used cooking oil recycling is the process of collecting oil left over from frying and turning it into useful products, such as biodiesel or sustainable aviation fuel, instead of pouring it down the drain.

How does Noovoleum pay people for used cooking oil?

Users drop their used cooking oil into a UCOllect box, and the amount is credited instantly to their UCOllect app account.

Which AWS services power the UCOllect boxes?

The boxes use AWS IoT Greengrass at the edge and AWS IoT Core, Amazon MSK, Amazon ECS on Fargate, Amazon Timestream for InfluxDB and Amazon Managed Grafana in the cloud.

How does the AI assistant keep costs low?

It routes simple questions to the lightweight Claude Haiku model and only sends complex questions to Claude Sonnet on Amazon Bedrock, so each question uses the most cost-effective model that can answer it well.

Build your next AWS solution with CASA5

Whether you run connected devices in the field or want to put generative AI to work for your customers, CASA5 can help you move from idea to working solution on AWS, fast.