Page 576 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Partners: Grab, Gojek, Lazada, KB Capital, Woori Finance, Sinarmas Hana Capital, HSBC,
Accial Finance, E-2W OEMs & Battery-swap Operators

2�2 Benefits of the use case

The project delivers significant benefits by financing electric two-wheelers (E-2Ws), which
effectively displaces petrol bikes. This shift not only raises the share of clean energy in the
transport sector but also cuts per-kilometer energy use, contributing to a more sustainable and
energy-efficient transportation system. By deploying locally built IoT and AutoML technology,
the project modernizes financial and mobility infrastructure, ensuring that these systems are
upgraded for sustainability. Additionally, the project enables low-emission delivery and ride-
hailing fleets, which helps to lower particulate emissions in dense urban areas. This not only
improves air quality but also enhances the overall sustainability of urban transportation, making
cities healthier and more livable.

2�3 Future Work

To further enhance our initiative and expand its impact, we plan to extend our pilots to Vietnam
and the Philippines, adapting our models to align with local riding patterns and regulatory
contexts. This geographic expansion will ensure that our solutions are tailored to meet the
specific needs of these regions, maximizing their effectiveness and relevance.
We will also integrate battery-swap station data, including cycle counts and dwell time, into
our systems to refine residual-value estimates. This integration will provide more accurate and
reliable data, enhancing our ability to manage and optimize battery usage and performance.

To improve efficiency and resilience, we will deploy lightweight AI models on IoT gateways,
enabling sub-100 ms scoring and ensuring offline capabilities. This edge AI scoring will
significantly enhance our operational speed and reliability, even in areas with limited connectivity.

We are introducing Environmental, Social, and Governance(ESG)-linked loan products, offering
rate discounts tied to emission-reduction milestones verified by telemetry. This innovative
approach not only promotes environmental sustainability but also incentivizes users to adopt
greener practices, contributing to a more sustainable future.

To enhance our feature set, we will partner with telecommunications companies and ride-hailing
firms to incorporate telecom spend and trip-density features. These additional data points will
improve the predictive power of our models, providing more accurate and actionable insights.

We will also upgrade our models by incorporating self-supervised contrastive pre-training on
unlabeled telemetry data. This approach will boost accuracy even with sparse ground truth
data, enhancing our ability to make precise predictions and decisions.

To support these advancements, we will need additional resources, including more Graphics
Processing Unit(GPU) nodes for weekly retraining, expanded Long-Term Evolution(LTE)/5G
bandwidth, and the addition of two senior data scientists to our team. These resources will
ensure that we can continue to innovate and improve our services, delivering greater value to
our users and stakeholders.

540