Agentic AI Reshapes Investment Research: MaxQuant Users Streamline Heavy Workloads

High entry barriers, low transparency, and limited asset coverage often constrain traditional financial trading. Individual investors and beginners must possess coding and data engineering skills for implementing strategy design and execution. Moreover, even seasoned investors may deviate from optimal decisions when interpreting complex news and market signals due to information asymmetry. MaxQuant views these structural issues as reducing trading opportunities, restricting creativity and capping profit potential.

To address these issues, MaxQuant turns to agentic AI to bridge trading asymmetries, explore more open and efficient quantitative trading models, and expand financial inclusion with services accessible to ordinary users. However, agentic AI often operates in fragmented environments. Different agents run with distinct contexts, dependencies, and invocation protocols. Building such a system from scratch would require significant time and resources for orchestration, fault tolerance, and monitoring. In addition, financial trading is inherently volatile. Systems must scale during peak hours while satisfying strict compliance and auditability requirements specific to financial scenarios, posing substantial architectural challenges.

Most critically, the market lacks an agentic AI infrastructure purpose-built for financial trading. Many existing solutions focus on models while neglecting engineering scalability, or provide insufficient compute resources for multi-agent orchestration and execution. In addition, agentic AI infrastructure evolves rapidly. For an early-stage firm like MaxQuant, investing heavily in building, upgrading, and maintaining foundational environments would require significant dedicated resources. This would inevitably slow product iteration and prevent MaxQuant from standing out in the rapidly evolving competitive environment. To stay ahead, MaxQuant urgently needs elastic and compliant agentic AI infrastructure to support multi-agent operations efficiently.

Since the launch of Amazon Bedrock, MaxQuant has closely followed its development and updates. At AWS Summit New York 2025, Amazon Bedrock AgentCore was officially announced. MaxQuant immediately recognized that AgentCore's runtime capabilities could integrate seamlessly with its existing products and services, addressing engineering challenges such as multi-agent orchestration and unified context management. What's more, the elasticity and compliance capabilities of AWS services meet the requirements of MaxQuant for financial trading. After consultations with AWS architects and engineers,

MaxQuant decided to migrate all workloads to AWS to build a next-generation, agentic AI-driven platform for investment research and trading decision making.
 

Amazon Bedrock AgentCore provides a complete set of capabilities to deploy and efficiently run AI agents. It helps developers bridge the critical gap between proof of concept and production deployment, enabling the rapid, secure, and large-scale deployment and operation of AI agents. Amazon Bedrock AgentCore now provides seven core capabilities for AI agents, including runtime, memory, identity verification, gateway, code interpreter, browser tools, and observability dashboards.

In its quantitative trading system, MaxQuant first experimented with Amazon Bedrock AgentCore's runtime capabilities for basic tasks such as data parsing and strategy backtesting. Compared with traditional architectures, you do not need to maintain the underlying runtime environment when you use Amazon Bedrock AgentCore. Amazon Bedrock AgentCore offers greater flexibility to meet compute demands in high-frequency trading scenarios. As the collaboration deepened, MaxQuant progressively migrated more complex workloads onto other AgentCore modules, such as generating fine-grained customer profiles before trades or automatically optimizing multi-strategy portfolios during trades. With agentic AI for collaborative task execution like data ingestion, feature extraction, model invocation, and strategy evaluation, an automated loop from analysis to execution is built. This allows R&D efforts to focus on strategy development and user experience improvement. Furthermore, the developer team no longer needs to orchestrate AI agents when new agents are added. System maintainability and scalability have been greatly improved.
 

Besides using Amazon Bedrock AgentCore to handle complex multi-step reasoning and strategy optimization, MaxQuant also encapsulated lightweight logic — such as cache refresh, batch data cleanup, log processing, and state synchronization — into Amazon Lambda functions, triggered through Amazon EventBridge. This approach, which decouples heavy and light tasks, enables MaxQuant to focus on high-value business processing and better unlock the value of agentic AI.

MaxQuant leverages Amazon EventBridge to flexibly capture events from different systems and trigger the corresponding Amazon Lambda functions, implementing cross-module automation. Built on an event-driven architecture, Amazon Lambda provides inherent elasticity and isolation, scaling for lightweight tasks on demand in high concurrency conditions without blocking core workflows. From an extensibility perspective, when new features are required, the developer team only needs to add functions that can seamlessly integrate into existing workflows. This "change once, apply globally" approach significantly reduces O&M overhead and enhances overall system agility.

As a young startup, MaxQuant has benefited from AWS not only through the "AgentCore × Serverless × Event-driven Architecture" technology stack, but also through critical support along its growth journey:

• Technology: The AWS Solutions Architect team provided ongoing guidance, helping MaxQuant avoid early risks while integrating Amazon Bedrock AgentCore with a serverless, event-driven architecture. This allowed the developer team to stay focused on quantitative trading business logic and product experience. 
• Resources: Through the AWS Startups Program, MaxQuant received valuable cloud resource quotas, accelerating experimentation and product iteration.
• Expansion: AWS offers a platform for MaxQuant to connect with potential partners, provides clear pathways and guidance for listing on AWS Marketplace, and presents MaxQuant opportunities to engage a broader base of prospective customers and investors, thereby expanding MaxQuant’s market reach.

With AWS's global compliance and infrastructure footprint, MaxQuant is able to win customer trust during international expansion by lowering technical barriers and demonstrating its strong professionalism. These AWS supports not only enable MaxQuant to build technical stability and credibility in early stages but also create a virtuous cycle between R&D and business growth, accelerating its overall growth.
 

With Amazon Bedrock AgentCore and other AWS services, MaxQuant reduced the development cycle of its core business modules from 4 weeks to 1 week, and the overall development was completed in just 1 month, well ahead of the 12-month estimate. The new system delivers efficient capabilities for data parsing, strategy backtesting, customer profiling, and strategy portfolio optimization in quantitative trading.

Beyond accelerating development, AI adoption lowers costs: Using Amazon Bedrock AgentCore reduces the MaxQuant team from 8–10 to 2–3 members, saving USD 3 million in human resource costs. By adopting the pay-as-you-go billing mode, MaxQuant avoids GPU hardware investments and cuts infrastructure costs by 80%.

By leveraging AWS's industry-leading security and compliance practices, MaxQuant improved R&D and delivery efficiency, significantly reduced costs, and also reinforced security and trust in the financial sector. At a recent MaxQuant community gathering in London, numerous quantitative traders and strategists expressed strong interest in the system, further proving the system's technical breakthrough and significant value.

While gaining recognition, MaxQuant has also shared its perspective on the future of multi-agent applications in financial scenarios. Zhang Weitong noted: "Agentic Trader will be among the first agents capable of fully autonomous operation, delivering execution efficiency, coverage, and speed beyond human capacity. It will enable seamless collaborations across borders, institutions, and asset classes. Humans, in turn, will be freed from repetitive tasks, instead focusing on defining values and providing governance oversight. What MaxQuant validates is not only the return potential of agentic AI-powered trading strategies, but also the resilience and autonomy of agents to operate in complex

social systems and meet real-time compliance requirements. We look forward to partnering with like-minded organizations such as AWS to usher in a new era of financial intelligence."

Looking ahead, MaxQuant hopes to continue working with AWS to further deepen the application of AgentCore and explore differentiated paths for multi-agent systems in financial scenarios. Building on single-marker analysis, MaxQuant aims to introduce agentic AI to cross-asset portfolio and hedging management covering stocks, options, and digital assets such as real-world assets (RWAs). This marks the next step toward its vision of "building a one-stop quantitative trading system with AI agents."