Trumid: Revolutionizing Credit Trading & Real-Time Bond Pricing With AI

In the dynamic world of electronic credit trading, traders and portfolio managers require precise tools to efficiently analyze and interpret data for strategic decision-making.

Trumid, a financial technology company that has designed a leading fixed income electronic trading platform, aims to optimize the credit trading experience by combining agile technology and market expertise with a focus on product design.

Building a Real-Time Bond Pricing Model

As Trumid sought to set itself apart in a crowded financial marketplace, they were faced with the challenge of developing a reliable and objective, real-time bond pricing model.

The solution would help their clients to:

• Manage and process vast amounts of data, keeping pace with often unpredictable and complicated market dynamics
• Work effectively for less liquid securities with few or no trades
• Coherently adjust pricing to align with comparable securities and account for factors like broad market indices, interest rates, and credit spreads

In addition to accurate pricing, the system would also need to deliver comprehensive supplementary data including confidence scores, intervals, and bid-offer spreads. This additional detail would greatly help clients decide whether to rely on the model for automated trading or take a manual approach.

Dataiku as the Central Hub

In collaboration with Dataiku, Trumid was able to seamlessly integrate multiple data sources, including direct market observations, Trumid’s proprietary market data, and

publicly available and contributed data sets, to produce a fully automated, unbiased – and accurate – two-sided market. Trumid utilized Dataiku’s advanced feature engineering capabilities to process this data, turning it into inputs for their machine learning (ML) models.

Dataiku’s robust suite of ML features was also leveraged to continuously train models on this engineered data, validate hypotheses, and select the most performant models based on real-world performance. Trumid also used Dataiku’s model deployment functionality to rapidly transition their most effective models from research into production.

The result was a comprehensive solution that helped to overcome their development and data challenges: Trumid’s proprietary Fair Value Model Price (FVMP™).

Since implementing Dataiku, Trumid has found that productivity has increased as the integration of Dataiku and deployment of Trumid FVMP™ has delivered on what it was built to do — providing real-time, predictive bond pricing information, while simultaneously reducing the time spent on manual model QA.

Dataiku’s centralized system has also helped to augment transparency across all our data pipelines and ML models, boosting robust model governance and aiding early anomaly detection. Mutisya Ndunda Head of Data Strategy and AI, Trumid

Finally, Trumid found that the reliability of their ML models increased significantly, enhancing overall client adoption and operational efficiency.

The Strong Value of AI & ML Investment

The deployment of Trumid FVMP™ through Dataiku has created significant value for Trumid’s automated request-for-quote (RFQ) trading solution, Trumid AutoPilot™. Designed to ensure efficient trade execution, this new product uses predetermined parameters, set by a client, to allow the platform to seamlessly execute RFQ trades on their behalf. Trumid clients that are using this new workflow are already finding value with 81% of eligible Trumid RFQ line items executed “no touch” in Q4 2024.

Innovating with the future in mind has also brought opportunities for data commercialization. Trumid’s real-time pricing feed offers immense value to clients. Because of the real-time nature of the data, they’ve been able to help their clients in a competitive market, where timely information can make the difference between profit and loss.

Trumid’s clients have enjoyed the transparency that FVMP brings. With regular updates every 30 seconds — a significant 10x increase compared to their prior models that updated every five minutes — their users now have access to real-time bond prices, ensuring they are in tune with current market conditions.

The impact of these advancements extends well into the future of Trumid’s business as they explore the development of even more sophisticated, AI-based trading tools and products.

Overall, Dataiku’s platform brings the flexibility that Trumid needs to move efficiently through the stages of ML model creation, boosting operational agility.

By using Dataiku, our data processing and model training times have been significantly reduced. The platform’s advanced features and tools enable us to streamline complex tasks and generate real-time bond prices much faster. Mutisya Ndunda Head of Data Strategy and AI, Trumid

As more and more members of their organization have grown accustomed to using ML and AI solutions, they have been able to research a multitude of models at the same time. Dataiku has allowed them to manage numerous data sources in a wide range of ML model research use cases.

Lastly, Trumid has found that Dataiku’s governance features help them mitigate risk by providing transparency across all ML operations. Trumid can track the performance and evolution of models, helping to identify anomalies early, and increasing accuracy and reliability over time in their machine learning operations

Source: Trumid internal data.

About Trumid 

Trumid is a financial technology company and fixed income electronic trading platform focused on US dollar-denominated Investment Grade, High Yield, Distressed, and Emerging Market bonds. Trumid optimizes the credit trading experience by combining agile technology and market expertise, with a focus on product design. The result is a differentiated ecosystem of protocols and trading solutions delivered within one intuitive platform. Learn more at www.trumid.com.