May 2025 Summaries

The author of this text is a developer at MongoDB who experimented with conformance checking, a technique to ensure that the implementation matches the formal specification written in TLA+. The author and their colleagues tried two techniques: trace-checking and test-case generation. Trace-checking involves generating execution traces from the implementation and checking if they conform to the spec, while test-case generation involves creating tests based on the spec and checking if the implementation passes them. The authors' experiments with conformance checking were initially unsuccessful due to difficulties in snapshotting the state of a multithreaded program, discrepancies between the spec and implementation, and limitations of existing tools. However, they learned valuable lessons about the importance of ensuring that the spec and implementation conform from the start, addressing discrepancies promptly, and using more advanced techniques such as model checking and specification-level state exploration. The authors also note that conformance checking can be a powerful tool for catching bugs and verifying correctness in distributed systems, but it requires careful planning and execution. They highlight several research prototypes and papers that have improved conformance checking techniques, including the development of new TLA+ specs, test-case generation tools, and trace-checking features in TLC.

Conformance checking is a problem at MongoDB that involves ensuring the implementation of distributed algorithms conforms to their formal specifications. The company uses TLA+ (Temporal Logic and Applied), a formal specification language, to describe its algorithms. However, testing conformance has been challenging due to the complexity of the systems and the need for continuous synchronization between the spec and implementation. Two techniques used for conformance checking are trace-checking and test-case generation. Trace-checking involves generating execution traces from an implementation and verifying they conform to a specification, while test-case generation starts with a specification and checks its behaviors in the implementation. The authors of a 2020 paper experimented with these techniques on two MongoDB products but found that conformance checking was difficult due to the complexity of the systems and the need for continuous synchronization between the spec and implementation. However, they learned three lessons: it's hard to snapshot a multithreaded program's state, the implementation must conform to the specification, and test-case generation should extend easily to multiple specs. Since then, there has been progress in conformance checking techniques, including new tools and research prototypes that aim to make conformance checking more practical and accessible.

The author of this text, a MongoDB engineer, discusses the challenges of conformance checking in software development. Conformance checking involves testing that an implementation matches its specifications, which are often written in formal languages like TLA+. The author describes two approaches: trace-checking and test-case generation. Trace-checking starts with an implementation and checks if it conforms to a specification by generating execution traces and verifying that they match the spec's behavior. Test-case generation starts with a specification and generates tests to verify that the implementation follows its behaviors. The author shares their experience with these approaches, highlighting successes and failures, and discusses lessons learned. They also mention ongoing research in conformance checking techniques, including new tools and methods for improving the process. The goal of conformance checking is to ensure that code matches specifications, which is crucial for developing reliable and maintainable software systems.

MongoDB has announced two significant updates to strengthen its security posture: expanding its public bug bounty program to include the free tier of MongoDB Atlas and partnering with GitHub to detect credential leaks faster. The inclusion of MongoDB Atlas in the bug bounty program enables continuous, real-world testing by a diverse security research community, systems designed for faster detection of vulnerabilities, and stronger confidence in MongoDB's ability to safeguard sensitive data. Through its partnership with GitHub, MongoDB can quickly identify potential security risks related to exposed credentials and notify customers, reducing the risk of unauthorized access or data leaks. These updates reflect a broader commitment to helping developers and organizations build secure applications, detect risks early, and respond quickly to new and emerging threats.

A unified database architecture significantly reduces development complexity and improves data consistency compared to a split architecture. In a unified approach, a single database platform handles both operational data and vector search functionalities, simplifying the data model and eliminating synchronization requirements between separate systems. This leads to increased developer velocity, reduced time-to-market, and improved operational reliability. Unified architectures also eliminate issues like "ghost documents" and reduce the complexity of synchronization mechanisms. In contrast, split architectures introduce additional complexity, network latency, and operational overhead due to the need for synchronization between separate systems. However, a unified architecture may not be suitable for all use cases, particularly those requiring specialized search or vector database expertise. Ultimately, the choice between unified and split architectures depends on specific requirements and constraints, and organizations should evaluate their own architecture to ensure it enables them to ride the wave of AI innovation.

A unified database architecture significantly reduces development complexity by eliminating synchronization challenges between separate vector and operational databases. Data consistency is guaranteed through atomic transactions in unified systems, preventing "ghost documents" and other split architecture failures. The total cost of ownership is typically lower with unified architectures due to consolidated infrastructure and reduced maintenance burden. Developer velocity increases with unified approaches as teams can focus on building features rather than integration code and error handling. MongoDB Atlas provides future-proofing benefits with integrated AI capabilities like vector search, automatic quantization, and more. A unified architecture unifies operational data and vector search functionalities in a single database platform, simplifying the data model by storing embeddings alongside associated data in the same document structure. This approach aims to provide the best of both worlds – the rich functionality of a specialized vector store and the reliability/consistency of a single operational datastore. The unified architecture eliminates the need for synchronization, reduces the risk of ghost documents or inconsistent reads, and offers simplicity, consistency, and speed of development. It is more future-proof as AI capabilities evolve at an accelerated pace, allowing for immediate leverage of new features. The choice of database infrastructure is pivotal in executing AI innovation, and a unified architecture is recommended to harness AI and data together seamlessly.

A unified database architecture with MongoDB Atlas provides a more streamlined approach to managing AI data, eliminating the need for synchronization logic and reducing operational complexity. In contrast, a split architecture involves separating vector operations and transactional data management into separate systems, which introduces additional complexity and potential inconsistencies. The unified approach offers benefits such as improved consistency guarantees, reduced latency, and increased developer velocity, making it a more future-proof choice for AI-powered applications. While a split architecture may be suitable for legacy systems or specialized requirements, the unified architecture with MongoDB Atlas is well-positioned to address emerging AI needs and provide a solid foundation for intelligent data management.

In the insurance industry, agentic AI is transforming claim processing by enabling autonomous systems to perceive, reason, and act independently. Insurers are investing in these technologies to deliver hyper-personalized customer experiences and capitalize on the growing $79.86 billion AI insurance market. AI tools like natural language processing, image classification, and vector embedding help insurers effectively manage claim-related data, generating precise catastrophe impact assessments and minimizing financial losses. An agentic AI agent manages a multi-step claim processing workflow, analyzing policy information and related documents using large language models and storing the outcomes in an Atlas database. The agent's autonomy is crucial to its effectiveness, but requires careful balancing between reliability and prescriptiveness. By crafting a prompt that describes the scope and goals of the task, developers can help agents understand their context and operate effectively, even if they lack complete autonomy.

This issue of "Innovating With MongoDB" highlights the importance of adaptability in business success, particularly in today's fast-paced and ever-evolving world. It showcases how organizations like LG U+, Citizens Bank, BioIntelliSense, and Enpal are leveraging MongoDB's flexible schema, seamless scalability, and fully integrated AI capabilities to adapt to change and build agile foundations for real-time innovation and dynamic problem-solving. These customers have achieved impressive results, such as increased resource efficiency, reduced processing time, and improved security, by harnessing the power of MongoDB's vector search, time series collections, and generative AI capabilities. The stories demonstrate how MongoDB empowers businesses to innovate and stay ahead of emerging trends and challenges, providing a platform for real-time insights and data-driven decision-making.

Voyage introduces its new embedding models, voyage-3.5 and voyage-3.5-lite, offering enhanced retrieval quality at the same cost as their predecessors, voyage-3 and voyage-3-lite. These models support embeddings in multiple dimensions, utilizing Matryoshka learning and quantization-aware training, which enable a range of quantization options. The new models outperform OpenAI-v3-large in retrieval quality while significantly reducing vector database costs. Voyage-3.5 and voyage-3.5-lite, evaluated across 100 datasets in diverse domains including technical documentation, code, law, and finance, show superior performance in retrieving documents based on cosine similarities. Additionally, the models are available through the Voyage AI APIs and the private preview of automated text embedding in Atlas Vector Search, providing a cost-effective and high-performance solution for various applications.

Ojje is a startup that aims to revolutionize literacy education by providing an adaptive learning platform that offers stories in 15 different reading levels, available in both English and Spanish, to cater to varying reading abilities within a single classroom. The company's mission is to reach every learner by offering materials that are leveled, accessible, and engaging, with the goal of vastly improving reading outcomes across K-12 education. Leveraging groundbreaking technology and partnering with MongoDB, Ojje has developed a dynamic platform that adapts to individual students' needs, allowing educators to personalize instruction and connect with every student. With plans to expand its library to include STEM materials and introduce tools for educators to track each reader's progress in real-time, Ojje is poised to address future literacy challenges and make literacy more accessible, engaging, and joyful for all students.

MongoDB Atlas is now available as a Microsoft Azure Native Integration, allowing enterprise customers to easily create and manage MongoDB Atlas organizations while consolidating billing for Atlas within the Azure console. This integration enables developers to interact with MongoDB Atlas as if it were a first-party service from Azure, providing streamlined and fully integrated experience. With this integration, organizations can harness real-time insights, scale globally, and accelerate their AI-driven roadmaps, unlocking new possibilities for innovation and transformation. The integration also reduces technical complexity, offering integrated billing, consolidated support, and simplified deployment, while enabling seamless operations and accelerated development.

The automotive industry is evolving rapidly, with drivers expecting more intelligent and personalized experiences. To meet these expectations, automakers are turning to generative AI-powered voice assistants that can answer questions, offer insights, and adapt to each user's preferences. However, building such intelligent systems requires a flexible and scalable data infrastructure, which MongoDB Atlas provides. By combining gen AI with powerful data infrastructure, companies like BMW, Honda, and HARMAN are pushing the boundaries of AI-driven car assistants. With MongoDB Atlas as their data layer, these companies can accelerate digital transformation, deliver next-gen in-car experiences, and unlock new levels of innovation. The solution is built on a flexible architecture that combines the power of MongoDB Atlas with Google Cloud's AI capabilities, enabling seamless, scalable, and secure deployment of gen AI apps.

OrderOnline, a social commerce platform in Indonesia, has improved its conversion rate by 56% after adopting MongoDB Atlas. With the help of MongoDB's flexibility, scalability, and AI capabilities, OrderOnline was able to manage unpredictable business processes, handle complex tasks associated with order management and logistics, and provide seamless search and customer support integrations. The platform built on MongoDB Atlas Search and Vector Search enabled features such as fuzzy search, synonyms handling, and AI-powered chatbots, which increased the success rate of deliveries and improved customer interactions.

The People Who Ship series brings behind-the-scenes stories from developers building production-grade AI applications using MongoDB. In this episode, John Ziegler shares insights learned while developing Central RAG, a centralized infrastructure that enables teams to rapidly build RAG-based chatbots and copilots for diverse use cases. Key takeaways include enforcing access controls across all operations, anchoring evaluations in the problem being solved, and educating users on what's possible and what's not. The backbone of Central RAG is the Credal tool, which provides a platform for teams to create AI applications on top of their data.

The new Pre-Migration Analysis feature in MongoDB Relational Migrator automates the evaluation process of complex database migration projects, providing a detailed and customized report that highlights inconsistencies, flags potential issues, and recommends actionable steps to ensure a successful migration. This feature empowers users to plan and execute migrations with confidence while saving time and minimizing risk. It minimizes disruption by providing recommendations for success, accurately allocates resources, accelerates the process and decision-making, reduces technical debt, and strengthens business cases. The tool connects to the relational database, extracts its structure, applies automated rules to identify potential migration issues, and generates a detailed migration risk assessment report complete with actionable recommendations. It also provides a "traffic light" migration confidence level indicating the overall readiness of the migration.

Capgemini and MongoDB are partnering to help businesses modernize their infrastructure, unify their data platforms, and power AI-driven applications. Many enterprises face challenges in adopting AI due to outdated infrastructure, including data fragmentation, lack of AI-ready data infrastructure, and difficulty integrating AI models. To address these challenges, MongoDB provides a flexible document model that enables seamless storage and querying of structured, semi-structured, and unstructured data, while its vector search capabilities enable semantic search, recommendation engines, and anomaly detection. Capgemini plays a crucial role in this transformation by leveraging AI-powered automation and migration frameworks to help enterprises restructure applications and transition to AI-ready architectures like MongoDB. Real-world use cases of their collaboration include AI-powered field operations for a global energy company, AI-assisted anomaly detection in the automotive industry, and making insurance more efficient through advanced data analysis. By modernizing their infrastructure and integrating AI-driven capabilities, businesses can move beyond legacy limitations and power the next generation of AI applications.

This article features Professor Chanda Raj Kumar, an Assistant Professor at KLEF Deemed to be University in Hyderabad, India. He is a MongoDB Educator and Leader of the MongoDB User Group—Hyderabad. With ten years of teaching experience, he has successfully empowered students to gain hands-on experience with MongoDB in their projects. His dedication earned him the 2024 Distinguished Mentor Award from MongoDB. The professor shares his educational journey, which consists of an undergraduate degree and M.Tech, and his current pursuit of a PhD in machine learning. He is intrigued by how efficient storage mechanisms impact data retrieval and accuracy, leading him to explore MongoDB's flexible JSON-based document structure. Professor Kumar teaches database-related courses across different levels, including Database Management Systems (DBMS), Python Full Stack, MERN Stack, and Java Full Stack Development. He incorporates MongoDB into his curriculum to prepare students for the growing demand for NoSQL technologies. The professor has built an active student community around MongoDB on his campus by incorporating hands-on projects, group discussions, and special training sessions called "Skill Weeks." His teaching approach is structured, starting with basics and gradually moving to advanced topics. He emphasizes hands-on learning, real-world projects, and guided problem-solving. The professor credits the MongoDB Academia platform for supporting him in enhancing his teaching methodologies and upskilling his students. Through surveys and project reviews, he has identified key insights about students' experiences with MongoDB, including strengths and weaknesses. These insights have helped him refine his teaching style. One memorable experience was during Skill Week, where 80% of his students earned MongoDB certifications. The professor's role as a MongoDB Educator has impacted his professional growth and the growth of the student community at his university, providing industry exposure and breaking the stigma about NoSQL databases. He advises educators to build upon students' pre-existing knowledge while introducing the transition to NoSQL concepts, highlighting key differences between SQL and NoSQL, and explaining when to use each.

The text discusses the role of Agentic AI in transforming investment portfolio management. Traditional systems struggle to process vast amounts of data from real-time market feeds, social media data, and other sources, making risk management increasingly complex for investment portfolio managers. AI agents, which can operate autonomously and take actions based on goals and real-world interactions, are set to revolutionize how portfolios are managed. MongoDB empowers AI agents to analyze vast financial datasets, detect patterns, and adapt in real-time to changing conditions dynamically. The solution includes a data ingestion application, three AI Agents (Market Analysis Agent, Market News Agent, and Market Assistant Agent), and a market insight application that work together to create a more intelligent, insights-driven approach to portfolio management. By leveraging MongoDB's flexible schema, AI agents can analyze diverse datasets, including unstructured alternative data like financial news and social media sentiment, and provide actionable insights for alpha generation and higher portfolio performance. The future of investment portfolio management is not about AI replacing human portfolio managers but rather humans and AI working together to create more intelligent, adaptive, and risk-aware portfolios.

Risk management in capital markets is becoming increasingly complex for investment portfolio managers due to the need to process vast amounts of data from real-time market feeds and unstructured social media data. AI agents, a type of artificial intelligence that can operate autonomously and take actions based on goals and real-world interactions, are set to transform how investment portfolios are managed. According to Gartner, 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024. At least 15% of day-to-day work decisions are being made autonomously through AI agents. MongoDB empowers AI agents to effectively transform the landscape of investment portfolio management by leveraging the combination of large language models (LLMs), retrieval-augmented generation (RAG), and MongoDB Atlas Vector Search, enabling them to analyze vast financial datasets, detect patterns, and adapt in real time to changing conditions dynamically. This advanced intelligence elevates decision-making and empowers portfolio managers to enhance portfolio performance, manage market risks more effectively, and perform precise asset impact analysis. Intelligent investment portfolio management involves selecting, balancing, and monitoring a mix of financial assets to achieve a higher return on investment (ROI) while managing risk effectively and proactively. It requires the integration of diverse unstructured alternative data like financial news, social media sentiment, and macroeconomic indicators, alongside structured market data such as price movements, trading volumes, index, spreads, and historical execution records. The complex data integration presents a new level of sophistication in portfolio analytics, requiring a flexible, scalable, unified data platform that can efficiently store, retrieve, and manage diverse datasets. MongoDB’s flexible schema accelerates data ingestion across various data sources, including real-time market feeds, historical performance records, and risk metrics, supporting more intelligent decision-making and proactive market risk mitigation. AI-powered portfolio asset allocation has become a desirable characteristic of modern investment strategies, providing insights tailored to specific financial objectives and risk tolerances. This approach optimizes portfolio construction by recommending an alternate mix of assets while continuously assessing the evolving market conditions. Agentic AI foundation incorporates an orchestrator layer that manages task execution in workflows, enabling AI agents to operate fully autonomous or semi-autonomous with a human-in-the-loop (HITL). AI agents are equipped with advanced tools, models, memory, and data storage, leveraging both long and short-term contextual data for informed decision-making and continuity of interactions. The solution includes a data ingestion application, three AI Agents, and a market insight application that work in harmony to create a more intelligent, insights-driven approach to portfolio management. By leveraging AI agents and MongoDB, financial institutions can unlock the full potential of AI-driven portfolio management to obtain advanced insights that allow them to stay ahead of market shifts, optimize investment strategies, and manage risk with greater confidence.

Ubuy, a Kuwait-based e-commerce company, is expanding its global reach by offering products from over 180 countries through its app and website. To support this growth, Ubuy adopted MongoDB Atlas to address scalability and search performance issues. The company migrated from MySQL to a fully managed MongoDB Atlas environment, implementing both MongoDB Atlas Search and MongoDB Atlas Vector Search to improve search efficiency and relevancy. With these enhancements, Ubuy has seen significant improvements in search response time, customer engagement, and business growth, including reduced search response times to milliseconds, over 150 million search queries handled annually with improved relevancy, and higher engagement on product pages due to AI-enriched content. Looking ahead, Ubuy aims to optimize search by consolidating inventory visibility across multiple stores, further enhancing its position as a leader in cross-border e-commerce.

The MongoDB for Educators program is a free resource that provides curriculum materials, MongoDB Atlas credits, certifications, and access to a global community of more than 700 universities. Professor Chanda Raj Kumar, an Assistant Professor at KL University Hyderabad, uses MongoDB in his teaching to empower students with practical skills and inspire future tech talent. He has successfully built an active student community around MongoDB on his campus, which has had a significant impact on the students' learning experience. The professor's approach to teaching MongoDB involves a structured pedagogy that combines hands-on learning, real-world projects, and guided problem-solving. This approach has resulted in 80% of his students earning MongoDB certifications during Skill Week, and he has been recognized for his contributions to MongoDB education with an award. The professor believes that integrating MongoDB into courses can help educators provide a comprehensive, real-world learning experience that prepares students for industry use cases.

The MongoDB Model Context Protocol (MCP) Server is now available in public preview, enabling AI-powered development by connecting MongoDB deployments to MCP-supported clients. The server makes it easy to interact with data using natural language and perform database operations with agentic AI tools, assistants, and platforms. This open standard protocol has been gaining traction as a way for AI agents and diverse data systems to connect, unlocking new levels of integrated functionality. Developers can now leverage the intelligence of LLMs to perform crucial database tasks directly within their development environments, with access to recent and contextually relevant data. The MongoDB MCP Server is available out of the box in Windsurf, an AI code editor used by over a million developers, and provides streamlined workflows and accelerated application development.