At Bolt, Bucharest is not a satellite office. It's a place where engineers ship core systems that serve over 200 million customers and 4.5 million driver and courier partners across more than 600 cities worldwide. The work ranges from machine learning and large-scale systems to privacy-by-design identity platforms and new product bets. Engineers here go deep, stay curious, and grow alongside the products they build.

“Programming languages and frameworks are just tools,” says Radu Ștefan Neacșu from the Identity group. “Engineers should be able to solve problems regardless of tech stack.” It’s a mindset that runs through the teams in Bucharest: hands-on, open-minded, and focused on what really works.

Identity. Secure by design, seamless by default

Every Bolt experience starts with an account. Identity authenticates users across Rider, Driver, Courier, and other apps. It balances strong security with a login that feels effortless.

Radu Ștefan Neacșu is a Staff Engineer serving as Group Technical Lead at Identity. He’s been at Bolt for almost seven years, long enough to see the company grow from a fast-moving startup to a global platform, yet he still finds it as exciting as in year one. “Seven years in one company might sound like a lot,” he says, “but when there’s growth, you get to work on something new almost every year. It’s like joining a startup again and again.”

Radu Ștefan Neacșu, engineer at Bolt

“The challenge is simple to state and hard to execute,” Radu explains. “Push security too far and you lock out good users. Push convenience too far and you invite fraud.” Bolt’s Identity platform strikes a balance between security and convenience, stopping fraud without locking out legitimate users. It provides passkeys, two-factor authentication (2FA), and security notifications whenever an abnormal account usage pattern is identified. It augments decisions with analytics that reveal where friction is too high and risk is too low.

Identity also powers KYC and user insights. It helps us understand who is behind an account so that we can personalise the experience and block abuse. Privacy is built in from day one. Dedicated privacy and compliance partners review designs early. The group maintains shared tooling for consent, data export, retention, and deletion. “We built generic systems used across Bolt,” says Radu. “That keeps compliance efficient and consistent.”

Recent work shows the scope and the standard. The team migrated both Rider and Eater apps to use JWT tokens for all their API calls to the backend, resulting in an 80% reduction in system load and a 50% cost improvement. The project was delivered with 0 downtime, and the transition occurred seamlessly for users who were already logged in, with the option to roll back immediately to the old way in case of disaster.

Another project the team worked on was automating the end-to-end personal data export, a complex task in a large ecosystem like Bolt, involving multiple verticals, multiple actors per vertical, and tens, if not over one hundred, databases of different types, including MySQL, Elasticsearch, and S3. Hence, the team built a solution that provided the backbone for such exports, orchestrating all the steps and helping teams owning data to integrate flawlessly via data plugins. “It was satisfying to see more and more plugins being added weekly,” Radu adds. “A good signal that we built it well.”

For engineers seeking meaningful challenges and opportunities to grow, the Identity team offers both. “We build scalable authentication and KYC systems used across the company,” Radu says. “Most of the engineers are in the Bucharest hub. That means long-term ownership and plenty of interesting challenges.”

For Radu, that mix of technical depth and constant evolution is what kept him motivated for 6 years at Bolt. “You can see your work shaping how millions of people interact with Bolt every day. And that kind of impact never gets old.”

Geo. Reading the city through data

When Cristian Ursache joined Bolt a year ago as a Data Scientist, he was already familiar with the company’s products. Still, he wanted to see how the technology behind them actually worked. “From the outside, you can tell Bolt operates on a huge scale,” he says. “What drew me in was the chance to work on something that impacts how millions of people move through cities.”

The reality matched quickly. “You get hands-on from the first week. It’s not just reading docs or shadowing, you’re seeing how your work affects live systems almost immediately.”

Cristian Ursache, engineer at Bolt

Cristian Ursache now works in Geo, where the map layer meets machine learning. “We do data science with an extra dimension,” he says. “Not just rows and columns, but heavily unstructured data such as coordinates, routes, regions, and time, which makes the problems and solutions more complex and beautiful.” The team processes signals from millions of rides per day. It learns how people move through cities and feeds that understanding back into the product.

One recent project predicts where a rider is likely to go next. We generate candidates from a rider’s history and local hotspots, then rank them using a learning-to-rank model based on gradient-boosted trees, utilising signals such as time of day, day of week, current origin, recency, and frequency of visits, as well as city popularity.

The model is trained on historical sessions paired with their corresponding candidate sets, and a session is considered positive when the trip ends at one of those candidates. Offline, we optimise a ranking score that rewards placing the right content near the top, while online, we monitor click-through rates on suggestions and track a practical top 3 relevance rate.

To make this work in production with low latency, we maintain an optimised, precomputed data structure of various features per pair of users and candidate destinations such that, when the Bolt app is opened, the model can fetch what it needs and return a ranked list quickly. In this way, we can surface likely destinations first, helping riders type less, start trips sooner, and have a home screen that is more closely aligned with their intent.

A representative example of the production ML stack is Bolt’s ETA prediction platform, which layers a learned correction on top of the routing baseline, updates estimates with live speed signals and historical patterns, and calibrates city by city, all while running continuously under millisecond response budgets. “Fast and fresh inputs matter,” Cristian says. “They keep search responsive, suggestions relevant, routing sensible, prices realistic, and dispatch timely.”

Geo also invests in better maps. The team built a workflow that lets trusted drivers report tricky spots or temporary closures. ML clusters and filters reports, then feeds updates back into routing and pickup suggestions. The loop improves as drivers see results and continue to contribute.

Culture makes this scale sustainable. “You will find colleagues who go deep on low-level details working alongside people who think in UX and product terms,” Cristian says. “Ownership is clear, check-ins are quick, and there is enough autonomy to move without bureaucracy.” The work is built in Romania and designed for global use.

Delivery. New products, real impact

When Dan-Mihai Dinu joined Bolt over two years ago, he had already worked in Microsoft Teams, Grafana and Skype, but wanted something more tangible. “I always wanted to be in the service of people,” he says. “Now I help restaurants reach more customers, couriers earn more, and people get meals easily. It keeps the city running a bit more smoothly and sustainably.”

Dan Mihai Dinu, engineer at Bolt

As an Engineering Manager, Dan leads teams that build and scale new product lines for Delivery. He remembers being surprised by how hands-on the work was. “People often imagine Bolt as a big, polished company where processes slow you down,” he says. “The reality is the opposite: you can still move fast, experiment, and ship things that reach millions in days or weeks.”

He’s also quick to challenge a few myths about Bolt in Romania. “Some engineers think that once you join a global company, you’ll just maintain existing systems or do ‘outsourced’ work,” he says. “But here, the teams in Bucharest actually build and own the core systems. We make the decisions, design the architecture, and see the full, global impact of what we ship.”

His teams ship visible features and behind-the-scenes engines. They launched DineOut, a new business line that helps restaurants fill tables by connecting with customers who eat in, not only those who order delivery. They also automated sales flows that used to be manual. That makes it possible to launch in new cities without growing headcount at the same pace. Efficiency and scale go hand in hand.

Machine learning and experimentation are built into Bolt products. Across the company, hundreds of models run in production, forecasting ETAs, optimising dispatch, powering search and recommendations, and supporting Alfred, the GPT-based customer support agent. Vision models read menus, while others verify driving licences and flag misuse. “We used machine learning long before it was cool,” Dan says. “Now we are pairing it with the new wave of AI to go even further.”

Dan Mihai Dinu presenting The Next Generation Engineer

Reliability is a habit, not a slogan: teams run on Kubernetes with strong observability, roll out changes gradually, and run safe A/B tests with guardrails and instant rollbacks. They learn from incidents. “We had a ‘Valentine’s Day saga’ three years in a row — there was always an incident on this day,” Dan smiles. “This year we finally broke the streak.” The takeaway is the same: build safeguards, react quickly, keep your sense of humour, and improve the system.

What keeps people here in the long term is the rare balance. “Bolt combines stability and innovation,” Dan says. “The foundations are strong. The culture is healthy. You can build a career and still move fast. I cannot think of another place in Romania where you do this at this scale and see your work used by millions.”

Why Romania, why now

Bucharest is one of Bolt’s largest tech hubs. That's not an accident. It reflects the quality of the engineering community and the impact the office delivers. “It is a sign the company values the quality of our work,” Radu says. “You can build world-class products without leaving the country.” Cristian adds a broader view: “Because Romania is one of our core markets, it's a great place to test ideas locally and scale them globally.”

Bolt’s promise to engineers in Romania is simple. Drive and own real Research and Development-level work that impacts 200M+ customers. Ship on a global scale, learn from talented peers, and grow in a stable environment that takes privacy, safety, and reliability seriously.

If you want ownership and impact without giving up stability, work on systems that matter and build from Romania, reaching the world, you will feel at home here.

Explore open roles on the Bolt careers page and build your next chapter with us.

Building world-class tech in Romania. Inside Bolt’s Identity, Geo, and Delivery teams

We’re hiring!

Recent posts

SUT Lab launches Asunción transport accessibility study