How crucial is AI in enhancing mobility today?

The fundamental premise of shared mobility is all about creating a highly scalable marketplace of riders and earners. Riders are looking for reliable, cost effective, safe and personalised ways to move around, and earners are aiming to optimise their time and earnings along with quality of work. Advancements in the realm of network speed and computational capabilities being as they are, it is now possible to gather, process, learn and recommend patterns at scale which can drastically improve mobility landscape. AI essentially deploys techniques to work on large-scale data to make informed real-time decisions. The ability to make such decisions at blitz-speed is transformational for mobility in many ways.

Some examples: 

• Predicting and forecasting demand patterns to plan optimal supply for mobility needs of any market.
• Predict ETA (Estimated Time of Arrival) and ETD (Estimated time to Destination) for the trips with high confidence and accuracy.
• Learn individual consumer habits and preferences and recommend mobility options that are personalised based on the history of trips and situational needs.
• Increasing the bar on safety by listening and reacting to events in near real time while continuously learning from such patterns.

Much like the diversity in each city, their mobility infrastructure, rules and regulations are fundamentally distinct from each other, too. There is no other way but to use Machine Learning and AI at scale to deliver Uber-grade experience across all mobility markets.

(Manish Bharani, Manager - Engineering, Uber India)

How is Uber leveraging AI to make mobility smarter and more intelligent?

Uber is a hotbed of innovations when it comes to Machine Learning. At any given time, our systems are dealing with orders or magnitudes of data flowing into our systems and decisions are being taken over it. At Uber, we have embedded AI in many aspects of our mobility business to make trips safer, reliable and highly personalised for everyone and everywhere. Needless to say, data is processed strictly as per extant laws of the countries we operate in.

Here are the few examples

• Improved Location Accuracy : To calculate vehicle location, Uber’s platform uses Global Navigation Satellite System (GNSS) signals coming from the receiver on the driver’s phone. Accuracy of these signals can be compromised in crowded urban places. Use of powerful AI based robotics algorithm and blending of other device signals, Uber has been able to improve location accuracy that results in better ETAs
• Maps Intelligence with Destination Prediction : Maps at Uber deploy context-aware suggestions for destinations, taking into account the rider’s current location, time of request, and historical information. Even for users that are new to the platform or a given city, we provide destination suggestions using aggregated information and the conditional probability that any user would select a particular destination given the context of time and space.
• Enhancing Marketplace intelligence : Uber uses Michelangelo, our ML and event processing platform for forecasting, Dispatch, Personalisation, Demand Modeling, and Dynamic Pricing. We build and deploy ML algorithms to handle the immense coordination, hyperlocal decision making, and learning needed to tackle the enormous scale and movement of our transportation network.
• Forecasting Demand and Supply : Forecasting supply and demand situations within cities can be done at scale with advanced modelling techniques. This has helped in building contextually aware and personalised products. 
• Fraud Detection : AI-based fraud detection helps in recognising patterns in transactions which are flagged as potential fraud and triggers multiple workflows for additional scrutiny of riders to provide safe
• Safety : Many AI-driven innovations in this space such as face recognition tech, document scan etc are helping make rides safer for Uber. Our advanced Selfie tech helps detecting authenticity of our driver partners and has also been helping to detect use of masks with riders and drivers during COVID times.
• Smarter Customer Support : Uber deploys Customer Obsession Ticket Assistant (COTA), a tool that uses machine learning and natural language processing (NLP) techniques to help agents deliver better customer support. Leveraging our Michelangelo machine learning-as-a-service platform on top of our customer support platform, COTA enables quick and efficient issue resolution for more than 90 percent of our inbound support tickets. 
• Driver / Rider One Click Chat (OCC) : Conversational AI advancements provide for a new smart reply feature called one-click chat (OCC). With OCC pre-trip coordination between riders and driver-partners is faster and more seamless. Leveraging machine learning and natural language processing (NLP) techniques to anticipate responses to common rider messages, Uber developed OCC to make it easier for driver-partners to reply to in-app messages.

In India, what are the biggest challenges in mobility and transportation, and how is Uber leveraging emerging technologies to address these challenges?

It is significantly challenging to predict the marketplace in the hyperlocal Indian scenarios. Road blocks, traffic diversions, congestion created due to people activities, and events are what differentiate Indian roads from other countries. Uber uses its Hexagonal Hierarchical Special Index Tech ( H3 ) to index cities as fine grained neighbourhoods for accurate marketplace predictions. H3 is used throughout Uber to support quantitative analysis of our marketplace, and now that it’s open sourced, you too can hexagonify the world!

Unlike western countries, Indian riders have greater familiarity with POIs ( Point of Interest ) vs Google marked locations. Riders more often are likely to start or end their trips at a ‘popular’ corner shop or a historic monument nearby. Uber’s advanced caching technology powered by enhanced Pickup & Drop Off Tech ( ePuDo ) makes it possible by surfacing POIs as default choices while allowing to finetune these further with precise pickup markup pins.

Reservations tech at Uber now allows reserving cars in India spanning a few hours to the entire day thereby significantly increasing the reliability and flexibility to move around with the same driver and multiple places at will. Similarly - Scheduled rides option now available for Indian riders as well allows for pre-scheduling rides for upcoming days in advance.

Uber’s partnership with few transit ecosystems in India and using seamless payment tech to pay seamlessly between many modes is a positive step towards building multi-modal technology for India-specific use cases. Cash was made popular within the Uber ecosystem from India-first innovations. Many products at Uber especially in Latin America now deploy cash into their payments tech mix. Uber platforms now have cash payments deeply embedded as part of their payment platform. Taking this forward - an increasing number of Indian digital wallets including UPI payments have been deeply embedded into Uber’s payment ecosystem.

Can you tell me the scope of work being done out of Uber's engineering center in Bangalore to enhance mobility?

The Bengaluru center started with focus on increasing access to Uber Products in the large, new and underserved Uber markets. Focus was towards developing products like Uber Lite and Uber Bus for emerging economies where access to Uber products was a significant challenge due to large app size incompatible with the low end phones, patchy networks and app complexities.

Mobility teams in Bengaluru believe in building locally and shipping globally. Over the last year or so, the ‘mobility’ function in Uber Bengaluru has significantly expanded to include Third Party ( 3P ) mobility use cases that comprise Car Rentals, MicroMobility and Local Cabs / Street Hailable products. This has been a significant shift from building consumer products over 1st-party supply that Uber ‘onboards’ to its marketplace vs 3rd Party marketplaces that Uber ‘integrates’ to exponentially grow Uber presence in new markets. 

The Bengaluru mobility engineering team also hosts a number of other teams such as Global Data intelligence and Edge API Gateway tech that focus on building engineering platforms for mobility products to scale and work in many markets across the world.

What is the potential of 5G in enhancing your current suite of engineering solutions?

Improvement in network speeds will significantly favour the AI program at Uber. 5G speeds will help us to build richer User Experiences by enabling real time data processing and decision making in building context aware AI applications. Demand and Supply modelling for faster marketplace forecasting, Conversational AI and Fraud detection are few of many use cases that have potential to leverage 5G in a big way at Uber.

What are Uber's plans for a clean and sustainable future in mobility? How does AI play a key role in sustaining these plans?

Uber is committed to reducing personal car ownership on crowded Indian roads by providing reliable, on-demand and safe transport for everyone and everywhere. Our AI-driven tech has been helping towards creating a balanced marketplace for riders and earners. Our marketplace tech uses AI-driven forecasting at scale to predict user supply and demand patterns in a spatio-temporal fine granular fashion to direct driver-partners to high demand areas before demand surges, thereby increasing their trip count and earnings.

AI in our products is also making them more intelligent and personalised. For example, our innovations in product recommendation tech is being positioned for promoting bikes and scooters ( MicroMobility ) for short distance commutes when riders choose to ride < 5 mile. This will save cost for riders while promoting environment friendly ways to reduce avoidable congestion resulting in reducing carbon footprint.

Our location and map intelligence now helps in positioning accurate pickup and drop offs to the extent of accuracy of multiple PuDo (Pickup & Drop Off) points at the large venues (Airports, large building complexes, malls etc) resulting in highly reliable trips that in turn contribute towards reducing traffic snarls on already busy roads.

AI safety center tech would ensure rider safety at scale by continuously monitoring trips and detecting anomalies to alert authorities maintaining city law and order if anything seems off at per trip levels.