Machine Learning Engineer - Reinforcement Learning
About the job
- A Master of Science or Ph.D. in Computer Science, Machine Learning, AI, or a related field, or equivalent practical experience.
- Hands-on experience building and applying machine learning in production-grade settings, with a strong reinforcement learning component such as policy learning, preference/feedback optimization, or offline/online reinforcement learning pipelines.
- Depth in deep learning, sequence modeling, and generative models.
- Demonstrated impact through strong publications or a clear history of shipping impactful machine learning systems end-to-end.
- Experience with large-scale distributed training and large-scale data processing.
- Ability to lead ambiguous technical work from problem framing through reliable delivery.
- Build scalable systems for training and fine-tuning large generative models that produce realistic, informative driving behaviors for evaluation and scenario coverage.
- Implement and iterate on reinforcement-learning-style methods, including algorithms, reward/preference objectives, and training setups suited to high-fidelity, insightful behaviors in simulation-aligned workflows with a closed-loop evaluation mindset.
- Ship deep learning solutions, including large language models and vision-language models where appropriate, that improve human-led triaging, automate high-volume workflows, and support nuanced analysis of self-driving behavior to surface critical anomalies.
- Own production-oriented machine learning for fleet-scale assessment, including training, optimization, monitoring, and iteration of models used to judge performance across large real-world exposure.
- Design and evolve data and evaluation systems inspired by reinforcement learning from human preferences and related paradigms, turning preference/judgment signals into repeatable, scalable training and evaluation loops.
- Partner with teams such as Prediction, Planning, Research, and platform/engineering leads to deliver cross-cutting improvements with clear metrics.
- Background in autonomous vehicles, robotics, or complex simulation environments.
- Strong grasp of modern reinforcement learning and post-training techniques in large language models, diffusion language models, vision-language-action models, and video generation.
- Hands-on integration of simulation platforms with machine learning training and evaluation workflows.
- Fluency in Python and experience with frameworks such as PyTorch.
- Experience defining and operating metrics for complex, safety-critical artificial intelligence systems.
- Technical leadership in influencing stakeholders, aligning teams, and raising the bar for evaluation rigor.
About the company
Pony.ai builds Level 4 autonomous driving systems as a full-stack, vehicle-agnostic Virtual Driver that combines its own software and hardware. Its main offerings span Robotaxi (via the PonyPilot app in Beijing, Guangzhou, Shanghai, and Shenzhen), Robotruck (autonomous freight and platooning), and licensing technology for Personally Owned Vehicles. The company earns revenue from ride-hailing services, freight logistics, and licensing fees through automaker partnerships with companies like Toyota, Sany, and Dongfeng Liuzhou Motor. Its goal is to scale Level 4 autonomy by expanding deployment of robotaxi and robotruck services, broadening licensing partnerships, and advancing automotive-grade hardware and software for mass production.
Company Size
501-1,000
Company Stage
IPO
Headquarters
Fremont, California
Founded
2016
Get referred to pony.ai
See people who can refer or advise you
Simplify's Take
What believers are saying
- Q2 2026 revenue rose 68.8% to $36.2 million, led by robotaxi growth.
- August 2026 Uber expansion targets over 2,000 robotaxis across five European cities.
- September 2026 Zagreb and Doha pilots validate international commercialization before year-end 2026.
What critics are saying
- Pony.ai burned $44 million operating cash in Q2 2026 against $1.39 billion liquidity.
- Regulators control expansion; stalled approvals in Europe or China freeze fleet deployment timelines.
- U.S.-China scrutiny on Chinese autonomous vehicles can block Uber partnerships and overseas scaling.
What makes pony.ai unique
- Pony.ai runs fully driverless commercial robotaxis in China, Zagreb, and Doha by September 2026.
- Its joint deployment model with Uber and Verne cuts capital intensity and speeds market entry.
- Gen-4 Robotruck uses shared robotaxi technology, slashing hardware costs 70% before production.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Health Insurance
401(k) Retirement Plan
401(k) Company Match
Life Insurance
Paid Vacation
Parental Leave
Disability Insurance
Growth & Insights and Company News
6 month growth
0%1 year growth
0%2 year growth
0%Uber and Chinese autonomous vehicle company Pony.ai plan to launch a robotaxi service in London, expanding their partnership into Europe. The companies will begin testing Pony.ai's Gen-7 robotaxis in the city in the coming weeks. The announcement follows Uber's strategy of partnering with over 30 autonomous vehicle firms to capture market share in autonomous delivery and ride-hailing. Uber expects to offer autonomous vehicle trips in 15 cities globally by the end of 2026. Uber and Pony.ai first partnered in May 2025, initially focusing on the Middle East. They plan to deploy 2,000 robotaxis across Europe, starting with Zagreb, where they launched a service in August 2026 with partner Verne. Uber also plans a separate London robotaxi service with British startup Wayve.
Pony.ai adds London to Uber robotaxi push in Europe. Pony.ai said it is working with Uber to bring its seventh-generation robotaxis to London, with testing expected to start in the coming weeks, taking the partnership beyond Zagreb, where Pony.ai-powered cars are already bookable on the Uber app. "Pony.ai on Uber is coming to London," the Chinese autonomous driving company said in posts on X and LinkedIn on Thursday. "Testing is expected to start in the coming weeks, and we look forward to working with local and national government officials, and our partners, as plans progress to begin trials for self-driving vehicles under the UK's leading AV regulations," Pony.ai said on LinkedIn. The company did not disclose how many vehicles it will test, which company will operate them, when a commercial service could start or whether test rides will be offered to the public through Uber. Pony.ai described London as "another step" in its expanding global footprint "as we scale autonomous mobility in Europe." Crowded London market. Uber already offers autonomous rides in London with British startup Wayve, which began supervised rides on the Uber app on Sept. 3 using Ford Mustang Mach-E vehicles with a safety operator on board. Riders requesting UberX, Uber Electric or Uber Comfort can be matched with a Wayve vehicle at no extra cost, and approvals for fully driverless operation were still pending at launch. Baidu's Apollo Go began road testing in London on July 28 with Freenow by Lyft, with safety operators on board and public rides planned for 2027, subject to regulatory approval. Baidu also has a separate agreement to put its robotaxis on Uber's platform in markets outside the United States and mainland China. From Zagreb to 2,000 cars. Pony.ai and Uber first announced their partnership in May 2025, saying Pony.ai robotaxis would launch on the Uber app first in a key Middle East market later that year. That launch has not been announced, and the companies' latest joint statement, in August, still described the Middle East only as part of their plans. Pony.ai's commercial pilot in Doha, which began in December 2025 with safety operators on board, runs through the Karwa app of Qatar's national mobility provider Mowasalat, not Uber. Under that arrangement, Pony.ai supplies its Gen-7 autonomous driving system, Verne owns and operates the fleet, and Uber offers the service on its app alongside Verne's own. Verne launched paid robotaxi rides in Zagreb in April, using Pony.ai's system on the BAIC Arcfox Alpha T5 with a safety operator on board. In August, Pony.ai and Uber said they plan to deploy more than 2,000 Pony.ai robotaxis across Europe, expanding from Zagreb to four more cities that have not been named. "The next chapter for autonomous mobility is about moving from individual launches to repeatable commercial scale," said Sarfraz Maredia, Uber's Global Head of Autonomous Mobility & Delivery. The companies did not disclose financial terms, and said vehicle funding and ownership may differ by market, with local fleet partners handling day-to-day operations. Zagreb riders have been able to book the cars through Uber since August, when requesting UberX or Comfort, with a licensed operator behind the wheel. Pony.ai and Verne began fully driverless test rides with invited passengers on a 22-kilometre route between Verne's headquarters, a Zagreb business district and the city's airport on Sept. 10. By then, the Zagreb fleet had covered more than 200,000 kilometres and completed several thousand customer rides, according to the companies. Uber has since completed its planned investment in Verne. "I can confirm - yes Uber has invested in Verne," Verne CEO Marko Pejković wrote on LinkedIn, without disclosing the size of the stake. Pony.ai's numbers. Pony.ai's revenue rose 68.8% to $36.2 million in the second quarter, while robotaxi revenue jumped 691.2% to $12.1 million, and its net loss narrowed to $45.4 million. The company had 1,975 robotaxis at the end of June and said in May it aims to reach 3,500 vehicles and more than 20 cities worldwide by the end of 2026. It says its planned and potential deployments outside China exceed 4,000 vehicles, including the more than 2,000 under the Uber plan for Europe. Pony.ai says its Gen-7 system cut the bill of materials for its autonomous driving kit by 70% from the previous generation. In Shenzhen, each Gen-7 robotaxi averaged 338 yuan ($50) in daily net revenue and 23 orders a day in February, a level Pony.ai says puts the city's fleet at unit-economics breakeven. The Gen-7 cars in Zagreb use a Pony.ai domain controller built on Nvidia's DRIVE AGX platform. Pony.ai also works with other platforms, including a Luxembourg test programme with Bolt and Stellantis. Bloomberg reported in October 2025 that Uber planned to invest about $100 million in Pony.ai's Hong Kong listing, which neither company has confirmed. Uber's wider bet. Uber said on Aug. 5 that it had pledged more than $10 billion to expand its robotaxi network, mainly through equity investments in autonomous driving partners and support for fleet operations. At the time, Uber was offering autonomous rides in seven cities and said it expected to reach at least 15 by the end of 2026. CEO Dara Khosrowshahi said in February that Uber intends to be "the largest facilitator of AV trips in the world" by 2029.
Mercedes-backed Momenta targets Dubai and Europe for robotaxi expansion. By Thomson Reuters Sep 30, 2026 | 3:12 AM By Qiaoyi Li and David Dolan Sept 30 (Reuters) - Chinese autonomous-driving company Momenta Global aims to have thousands of robotaxis on the road next year as it looks to expand to Dubai and more cities in Europe, the head of its robotaxi business told Reuters. Founded by a former Microsoft researcher and backed by Mercedes-Benz, Momenta counts Toyota, and BYD among its partners. It has two main businesses: advanced driver-assistance software, which it sells to automakers, and robotaxis, which are still in an early stage with more than 100 vehicles deployed in three countries. The company is testing its robotaxis in five Chinese cities, as well as in Munich and Abu Dhabi. Momenta has been in talks with a "few" European cities for further expansion and plans to deploy vehicles in Dubai next year, Shuo Xie, the head of its robotaxi business, said on Tuesday, adding that Japan was also a priority. "The size of our robotaxi fleet will be in the... hundreds at the end of this year and then it will be several thousand at the end of next year," she said in an online interview. Momenta competes with Alphabet's Waymo and China's Pony.ai, among others, in robotaxis. In July the company raised around $751 million in an initial public offering in Hong Kong. But its shares have fallen some 45% from the IPO price, hit in part by caution about the outlook for Hong Kong-listed AI shares. Momenta also remains unprofitable as it spends heavily on research and development. It has been working with chip company XHeart to produce chips designed specifically for its autonomous-driving software. The cost of such chips should be "significantly lower" than those of equivalent computing power from Nvidia, Xie said. XHeart is now working on a next-generation chip, the X9, she added. "In a couple of years the robotaxis that I am going to deploy around the world, hopefully a lot of them will be hosted on that X9 chip." (Reporting by Qiaoyi Li and David Dolan; Editing by Clarence Fernandez)
Pony.ai has unveiled its new Level 4 autonomous heavy-duty truck developed with GAC Commercial Vehicle at IAA TRANSPORTATION 2026 in Hannover. The Gen-4 Robotruck is built on GAC's T9 battery-electric platform, with volume production scheduled to begin later this year for deployment in long-haul freight, dedicated-route logistics and port transportation. The vehicle features fully redundant systems for steering, braking, communications, power supply, computing and sensing. It includes nine lidars, three millimeter-wave radars and 13 cameras for 360-degree coverage. The autonomous driving kit's bill-of-materials cost has been reduced by 70% compared with the previous generation, whilst transportation costs per ton-kilometre are expected to fall by 30%. Pony.ai plans to expand its Robotruck business internationally, including entering European and Middle Eastern markets over the next two years.
Zagreb robotaxis begin fully driverless passenger rides. Pony.ai and Verne have begun carrying passengers without an onboard safety operator on public roads in Zagreb, five months after launching Europe's first commercial robotaxi service. * 10 September 2026 Chinese autonomous driving company Pony.ai and Croatian operator Verne have started fully driverless passenger testing in Zagreb. Invited passengers can now ride without a human operator in the driver's seat along a 22-kilometre route connecting Verne's headquarters and a major business district with Zagreb's Franjo Tuđman Airport. The companies plan to extend the driverless operating area over the coming months. The tests come after Verne launched what the companies describe as Europe's first commercial robotaxi service in April. That service initially operated with an onboard autonomous vehicle operator. The latest phase is the next step in the partnership between Pony.ai, Verne and Uber. Pony.ai supplies the autonomous driving system, Verne manages the vehicles and service, while Uber provides access to passengers through its ride-hailing platform. "Fully driverless passenger rides in Zagreb mark an important step in Pony.ai's international growth," said Pony.ai Founder and CEO James Peng. Verne's service has covered more than 200,000 kilometres and completed several thousand passenger rides since April, the companies said, with an average passenger rating of 4.7 out of five. Uber has already begun offering the Zagreb service through its app, initially with a licensed operator monitoring from the driver's seat. The latest phase gives Pony.ai a European deployment where its autonomous driving technology, local fleet operations and ride-hailing access are already in place. The company has been using a similar partnership model in its overseas expansion, combining its autonomous driving technology with local operators and mobility platforms. In China, Pony.ai already runs fully driverless commercial services in four cities. It said its autonomous vehicles have travelled more than 100 million kilometres globally, including more than 40 million kilometres without a human driver. The company is targeting more than 3,500 robotaxis globally by the end of 2026. Its overseas deployment pipeline includes more than 4,000 vehicles, including plans announced with Uber to deploy more than 2,000 vehicles across Europe. Pony.ai said the new rides are being conducted under Croatian regulations and the approvals required for passenger-carrying autonomous vehicle testing on public roads. The companies have not said when fully driverless rides will become available to the general public.