Meet the Team: AI Engineering

Bloomberg’s Artificial Intelligence (AI) Engineering group, a team of more than 350 AI researchers and engineers from around the globe, are constantly exploring state-of-the-art AI tools and technologies and considering their applications in the financial domain. This is a technology that Bloomberg has been using since 2009, when the company deployed its first machine learning-enabled solution, an analytic on the Bloomberg Terminal that provided customers with a way to understand the sentiment of news stories.

Today, the AI Engineering group is spread across New York City, London, Princeton, and Toronto. Its engineers build AI solutions using machine learning (ML), natural language processing (NLP), information retrieval (IR), speech, computer vision, time-series analysis, and LLMs to help extract, enrich, search, summarize, and analyze the ever-increasing volume of structured and unstructured information that Bloomberg’s clients use to make informed decisions. Bloomberg’s AI-enhanced solutions have a range of applications in news, finance, communications, and research across the firm. In developing them, Bloomberg’s AI experts collaborate closely with domain experts from the company’s Core Product, Data, Enterprise Data, Enterprise Products, CTO Office, and application development teams to inform model training, evaluation, and inference, as well as for idea generation and product development to evaluations.

Members of the AI Engineering group are also active in the broader academic community, collectively publishing over 100 papers in peer-reviewed journals over the last three years, as well as attending and speaking at conferences in AI, ML, NLP, and IR. Team members also host the Bloomberg Data Science Ph.D. Fellows and serve as committee members for conferences.

Let’s meet the Head of Bloomberg’s AI Engineering group, Anju Kambadur. Anju oversees four distinct areas that are each pioneering their respective subdomain within the field of AI. Prior to joining Bloomberg, Anju was a research staff member at IBM Research and has published peer-reviewed articles in the fields of high performance computing, machine learning, and natural language processing. He was recognized on Insider’s “AI 100” list of the top people in AI in 2023.

Tell us about your role as head of AI and what the department is responsible for?
I look at my role a few different ways. As a people manager, I hold myself accountable to ensure that we hire the best AI researchers and engineers to be our colleagues, and that they are mentored and feel safe, supported, and challenged. As a technology leader, I hold myself accountable to ensure that we are building bleeding-edge infrastructure and innovating; as you know, the field of AI is changing very rapidly. As an engineering manager, I hold myself accountable to ensure that we are delivering high-quality, sustainable products on time and on budget. Finally, as a representative of Bloomberg Engineering, I hold myself accountable for showcasing our company, its products and services, and its culture to an external audience.

Just like any other engineering group, we are responsible for building and deploying infrastructure, services and applications; being focused on AI, our services typically are backed by machine learned models. In this capacity, we work closely with the firm’s Core Product, Enterprise Data, Enterprise Product, CTO Office, and Data organizations. Our models are used for tasks such as extraction, enrichment, search, summarization, and time-series analysis.

I have four areas that report to me:

• AI Search Engineering focuses on a set of systems that together provide the next-generation search experience across the Bloomberg Terminal (e.g., news, research, command line search).
• AI Enrichment Engineering focuses on the systems needed to parse, understand, and enrich documents and communications with information important to Bloomberg users (e.g., entities, securities, fields, topics, sentiment, and offers).
• AI Platforms Engineering focuses on building the infrastructure and tooling used to accelerate the model development life cycle (MDLC) for AI-driven product development across Bloomberg (e.g., Bloomberg’s Data Science Platform, NLP libraries, and models trained and fine-tuned for use in financial applications).
• AI Finance Engineering focuses on core time-series prediction problems for Bloomberg. Some of the products this team collaborates on include Bloomberg’s leading evaluated pricing service, BVAL, Intraday BVAL (IBVAL) Front Office, and Bloomberg Dividend Forecast (BDVD).

What are some of the unique technical challenges your team tackles?
There are aspects to machine learning that are different from software engineering. We operate in lockstep with our Product and Data counterparts, as we can’t build AI solutions without first having three things: a deep understanding of our clients’ objectives, labeled data to train our models, and data analysts and/or data management professionals to help evaluate our models. Second, we are dependent on CI/CD to keep our models and services fresh and relevant, as input data distributions change and the desired outcomes also change rapidly. Third, there is an aspect of upfront investment in compute and data that makes AI services more expensive to develop relative to traditional software services: we run on “different” hardware (i.e., GPUs) than most other software. Finally, all of us are watching one innovation after another in the AI space, which means that tech debt accumulates more rapidly in our field. So, we are constantly evaluating our tech stack and moving thoughtfully to more performant, yet more sustainable, infrastructure.

Briefly tell us about your career path?
I recently marked my 10th anniversary at Bloomberg. This is my second job. Prior to this, I was a research staff member at IBM Research in Westchester County, just north of New York City. Upon joining Bloomberg, I started as an individual contributor working on our NLP models, but was asked to change roles and be a team lead within a year of joining the company. Thanks to the leadership team in Product, Data, and Technology – including our CTO Office – we invested early and often in compute, data, and the talent needed to build products that had AI as a core component. With this investment, our team has grown rapidly – it is now nearly 30 times the size it was when I first joined it a decade ago.

What’s your strategy for choosing team members (for example, what skills, aptitudes, attitudes, experience, cultural backgrounds, gender/ethnicity, and/or abilities are you looking for)? / What skills do you look for when hiring engineers for your team?
By and large, we look for the same things in a potential colleague that Bloomberg’s Engineering department seeks: someone who is technically competent and a culture add. Since we specifically work in the field of AI, being “technically competent” means that we test for deep knowledge in disciplines such as machine learning, natural language processing, information retrieval, speech, computer vision, or time-series analysis. However, in the end, we are engineers and what we produce are engineering artifacts, so we test for the skills that allow you to build and maintain software. Similarly, since we work in teams, we seek people who will thrive in Bloomberg’s collaborative culture.

One thing worth calling out is that every AI service captures a small aspect of human decision-making (e.g., assigning a sentiment to an article about a company). So, in that sense, they are not just computations but truly capture culture. To us, diversity is an important aspect in ensuring that we build good services.

What are some of the factors driving the rapid growth/expansion of your team?
Decision-making in finance has always centered around drawing insights from data, so AI was a natural fit for finance. That is the main reason for our group’s growth over the past decade. Let me elaborate.

The rapid growth of AI in finance is due to some critical factors. Advancements in computing, such as GPUs, have enabled us to train large models as well as to process vast amounts of financial data, such as news, research, and exchange information. Advancements in machine learning, especially in deep learning, have made it possible to build AI systems that assist our clients in complex tasks, such as financial statement analysis. Cloud computing, open source code, and open weight models have also served as accelerants fueling the growth in use of AI. Plus, our clients’ appetite for AI applications and our company’s foresight to anticipate the future of financial workflows have also driven our team’s growth.

Today, we use AI to extract information from unstructured sources, such as news and research, and enrich it to help clients search for and discover content and insights on the Terminal and in our Enterprise offerings, in bond pricing and other such time-series predictions, and finally, in summarizing all this information using LLMs.

How do you foster culture on your team?
We have colleagues from more than 40 countries and it is important to have cultural practices that not only bind us as a group, but also ensure that the artifacts we produce have a high and consistent quality. In this sense, our culture begins with our hiring practices and how we conduct our interviews: together and not for any individual team. As soon as you become a member of the AI Engineering group, you’ll have a consistent onboarding experience that helps you understand how we get product requirements, how we codify those requirements into documentation, annotation guidelines, labeled data, modeling choices, evaluations, production rollout, etc. As for other cultural practices, we follow the Bloomberg Engineering culture. For example, coding standards and deployment practices.

With more than 300 people in the group, we are now too large for our culture to be propagated solely through in-person interactions. I now make use of surveys, semi-annual reviews, and also roll out programs to better define career development opportunities across our group. Our “young” organization has us in a unique and privileged position to create a market-leading team, a place where we thrive as AI experts, and most importantly, enjoy coming together because we have built an environment where everyone is valued, we challenge each other to continue growing, and we provide opportunities for continuous development. Culture is something we all contribute to, and this is how we are doing it in AI Engineering.

“Decision-making in finance has always centered around drawing insights from data, so AI was a natural fit for finance. That is the main reason for our group’s growth over the past decade.”

– Anju Kambadur

Edgar Meij is the Senior Engineering Manager who heads the aforementioned AI Platforms Engineering group. He’s been with Bloomberg since 2016, and has led multiple teams within the AI Engineering group, including AI Search Engineering. He was previously at Yahoo Labs in Barcelona, holds a Ph.D. in computer science from the University of Amsterdam, and has published more than 110 papers in top international venues.

Tell us about what you’re working on now and what your biggest challenge is. What inspires you most about it?
AI and its constituent fields, including ML, NLP, and search, are hot topics and that trend does not seem like it will be slowing down anytime soon. In recent years, we have made great strides in supporting AI applications across the firm, with tools such as the Bloomberg Data Science Platform, KServe, and Buildpacks, as well as other tools to help maintain, scale, and remediate model pipelines in production, such as our Data and Model Registry. In addition, for a long time now, we have offered highly-performant implementations of state-of-the-art NLP algorithms in a standardized way. More recently, we extended that offering to also include models, model architectures, components, and methods for training LLMs and other neural models. Benchmarking here is key, so we also provide unified interfaces for comparing base models across internal and external datasets and test collections.

We also offer our engineers an internal AI “Apps” environment where they can quickly stand up a UI as a proof-of-concept for their services. These tools enhance the productivity of our ML practitioners by enabling them to be more efficient and effective at all stages of the machine learning development cycle. One area we are explicitly focusing attention on this year is how we can better support workflows around data annotation, data curation, and user feedback. High-quality data is key to high-quality applications, as well as to keeping models up to date, so we need to make it easier to continuously gather feedback from our end users – in addition to our internal experts – and feed that information back into the model (re)training pipeline.

Last, but certainly not least, our customers are expecting increasingly smarter user interfaces and corresponding results. In order to support our engineers’ efforts to build these out, we help by investing in standardization around the use of LLMs, as well as in tools that support modern search, tool invocation, and stateful query answering capabilities.

Bloomberg has a distinctive culture. What attracted you to it?
That’s easy: its collaborative culture. Across the entire company, we are united in working towards creating the best experience for our customers, and giving them what they need to make financial decisions, no matter their context. This unity permeates throughout the organization and creates a unique identity and sense of community. In a similar vein, Bloomberg’s deeply ingrained culture of giving back through Best of Bloomberg (BOB) volunteer opportunities and our Corporate Philanthropy programs resonates deeply with me.

How do you foster a collaborative, inclusive environment at work?
I base this on three main pillars for my teams. First, I aim to actively get my team members to know each other and promote a sense of community through both social events and team-building BOB events. Second, I also champion a culture of understanding and respect by encouraging collaboration, communication, and continuous learning through attending conferences, hosting internal talks, and through the company’s various guilds. For instance, recent conferences, such as SIGIR 2024 and KubeCon + CloudNativeCon 2024, were great opportunities for attendees to meet each other in a different context than their normal day-to-day responsibilities. They also enabled them to learn from experts in the field and their colleagues, thanks to the numerous talks our engineers presented at the events. Last, but certainly not least, I support and promote our Diversity & Inclusion activities through recruitment, mentorship programs, and external events.

You’ve built a number of teams from scratch during your career at Bloomberg. What’s your secret sauce to successfully building a team and getting the team to gel?
I don’t claim to hold any secret sauce to building a team. However, what I have found works well is to be deliberate about defining your team’s vision, mission, and mandate and making sure it aligns with the company’s goals. Then, by aligning recruitment and hiring the best people for the team (not only technically but also culturally) – and with the right energy! – you create a flywheel effect.

Another thing I’ve learned is to be as clear as you can about expectations and responsibilities, and encourage ownership and innovation by explaining the “what” and “why” related to lines of work, but leave the “how” up to the team. Establishing clear lines of communication and giving feedback (both ways!) is paramount to continued success. And, of course, leading by example is still one of the best ways to inspire your team.

What else should we have asked you about? What else would you like to share about your Bloomberg experience?
If there is one thing that stands out from my interactions with colleagues across the firm, it’s that we are not only very hard-working, but also innately curious. My advice is to keep challenging assumptions, continue asking questions, and to speak up if that is the right thing to do.

“Across the entire company, we are united in working towards creating the best experience for our customers, and giving them what they need to make financial decisions, no matter their context.”

– Edgar Meij

Samuel Kola is a machine learning engineer working with the ML Productivity team within AI Platforms Engineering. He’s been at Bloomberg for almost five years, having first joined the firm’s Derivatives Pricing Engineering team after earning his master’s degree in data science.

Tell us about what you’re working on now and what your biggest challenge is. What inspires you most about it?
I’m spearheading the development of a pivotal API to centralize the sources of truth for our command line interfaces (CLIs), user interfaces (UIs), and SDKs. This centralized API will empower fellow engineers on the AI Platforms team to be able to seamlessly query global information addressing state, connectivity, and availability of our AI resources. The primary challenge lies in preventing this API from becoming a bottleneck or introducing a single point of failure, a delicate balance I’m committed to maintaining. What drives me in this project is the impact it will have on fostering agility in our systems.

It looks like you’ve worked on several Bloomberg teams throughout the years. What motivated your most recent moves?
I actively seek out new projects involving unfamiliar tools, approaches, and technology, thereby ensuring a continuous learning curve and personal growth. This has led me to move to a variety of different teams within Bloomberg.

What is it like moving to different teams within Bloomberg?
Team transitions are facilitated by a very supportive environment, with both one’s existing and new managers providing ample support, making the moves a smooth and collaborative process.

What surprised you most about them?
The most surprising thing I have experienced about team mobility was the unwavering support from my team lead in Derivatives Pricing Engineering during my move to the AI group. In addition to offering guidance, he also went above and beyond by offering to speak to the other managers on my behalf. This was an unexpected and greatly appreciated show of support.

You’re very involved in a variety of D&I initiatives. Tell us about why this is important to you and some of the things you’ve played a role in to make our workplace more inclusive.
Diversity and inclusion are vital to me, as they cultivate a rich tapestry of perspective, while driving innovation and collaboration. It’s about creating an environment where every voice is heard and valued.

I’m actively involved with the Bloomberg Black in Tech (BBIT) group’s mentorship and recruitment initiatives. I have played roles in implementing mentorship programs and advocating for diverse hiring practices. These efforts aim to break down barriers, promote equality and ensure everyone feels welcomed and respected in our workplace.

How does the collaborative environment at Bloomberg create opportunities to learn new skills and expand your expertise?
Bloomberg’s collaborative environment serves as a catalyst for continuous growth and development. At Bloomberg, I have been involved in cross-functional projects that have exposed me to a varied set of perspectives and skill sets, all of which have broadened my understanding of different domains. Examples include learning new programming languages, as well as about product and interface design.

Mentorship opportunities at Bloomberg have also been instrumental in enabling me to learn from other professionals and to improve both my technical and interpersonal skills. These initiatives, in addition to the vast number of training programs available at the firm, foster a culture of continuous learning and growth.

I am grateful for the collaborative spirit that defines our workplace, making it a truly exceptional environment in which to thrive and contribute.

“Bloomberg’s collaborative environment serves as a catalyst for continuous growth and development.”

– Samuel Kola

Jing Wang is a Team Lead for the News Enrichment team, part of the AI Enrichment group. She’s been with Bloomberg for five years, having started her career as a senior research scientist after receiving her Ph.D. in electrical and computer engineering.

Tell us about what you’re working on now and what your biggest challenge is. What inspires you most about it?
I lead the News Enrichment team, which provides a high-quality AI entity tagging system. This is used to correctly identify named entities, such as organizations, people, and products in news articles, and to link them to unique Bloomberg identifiers. This helps make relevant news on companies more discoverable so that clients can clearly understand what is happening with a company when they search for, monitor, or read relevant news stories. We aim to build scalable, domain-expert-in-the-loop AI systems that support multiple news sources and languages and uphold high-quality standards. In the meantime, we also need to focus on the robustness and reliability of the service so it can handle massive streams of millions of pieces of news content on a daily basis.

How do you foster a collaborative, inclusive environment at work?
Our team thrives on open communication, where every member has a voice that deserves to be heard. During the design and planning stages of our projects, everyone in the team participates and shares their opinions and ideas. Each member’s input is integral to the project’s success, leading to innovative and robust solutions. We also maintain a clear vision and set of goals that are communicated regularly, ensuring that every individual’s efforts are aligned with our collective objectives.

Team members are also encouraged to take ownership of projects. By entrusting them with greater responsibility, we’ve seen a marked increase in engagement and accountability. This has helped us motivate people to discover their potential and to be problem solvers. With people taking full responsibility for their work and projects, the team is driven to deliver high quality solutions.

What do you think a team must have to be effective and healthy?
It is crucial to always keep the bigger picture front-of-mind. It helps the team understand how their work contributes to the overall success of the company.

Focusing on the outcome, rather than the output, is another critical aspect. This approach ensures the team is aligned with the ultimate goal and not just engrossed in the tasks at hand.

Moreover, encouraging team members to acquire new skills and fostering a learning environment is vital for the team’s growth and adaptability. In our team, we have implemented a culture of continuous learning, in which team members are encouraged to upskill through various training programs and workshops.

“Our team thrives on open communication, where every member has a voice that deserves to be heard.”

– Jing Wang