Gradient - Your Finance OS 💸

CV Deep Dive

Today, we’re talking with Chris Chang, Co-Founder and CEO of Gradient.

Gradient is building an AI operating system for finance - designed to automate complex workflows and enhance decision-making across financial services, including asset management, banking, hedge funds, fintech, and more. The platform is powered by a finance-specific knowledge graph and expert AI agents that process vast amounts of financial data with high accuracy and low latency. By integrating AI into critical financial operations, Gradient aims to improve efficiency, reduce risk, and provide deeper insights for financial professionals.

Today, Gradient has been adopted by large enterprises and startups alike, who are all looking to scale their operations with AI while maintaining strict security and compliance standards. Gradient’s model-agnostic platform supports deployments across cloud and on-prem environments, ensuring data privacy while enabling real-time intelligence across the financial stack of an organization. 

In this conversation, Christopher shares the origins of Gradient, the challenges of scaling AI in finance, and the company’s vision for the future of AI-driven financial systems.

Let’s dive in ⚡️

Read time: 8 mins

Our Chat with Chris 💬

Chris, welcome to Cerebral Valley! First off, introduce yourself and give us a bit of background on you and Gradient. What led you to co-found Gradient?

My name is Chris, and I’m the CEO of Gradient. My background is a combination of finance experience and AI products. I started my career as an investment banker in restructuring advisory and later became a fundamentals investor before moving into tech.

For the last seven years, I’ve been an AI product leader. Most recently, I was running Studio AI at Netflix, where we applied AI to the content production lifecycle. It’s been exciting to use both my finance and AI expertise to solve problems in an innovative way.

How would you describe Gradient to the uninitiated enterprise or AI team?

We like to say that Gradient is the AI operating system for finance. It’s powered by our finance reasoning engine, which is a custom-developed AI designed to solve various tasks and problems within the finance space.

We work with a range of companies across asset management, banking, hedge funds, and fintech. The applications of this technology are diverse, enabling us to address different challenges across the industry.

Who are your customers today? Who is finding the most value in what you're building at Gradient? 

The core value we provide with our product is the ability to automate complex workflows within financial services and develop specialized co-pilots that are powered by AI agents to assist with high-skill knowledge work.

Across different companies, our use cases generally fall into two categories: front office and back office. On the back office side, we power systems for automated suspicious activity reporting for compliance and fraud detection, automate customer onboarding, manage document processing, and streamline the diligence process for transactions.

On the front office side, we provide additional intelligence for investors as they conduct diligence on potential investments, helping them to make more informed decisions.

Talk us through your Finance Reasoning Platform, which is a core pillar of Gradient’s product offering. If I'm an enterprise looking to automate elements of my financial workflows, where do I start? 

Our product is composed of two core technologies: our finance knowledge graph and our AI agents that specialize in all aspects of financial services.

The knowledge graph system we’ve developed is purpose-built to ingest a broad variety of financial data, which is naturally complex. This includes securities contracts, SEC regulations, internal policies, and market data. Our knowledge graph engine interprets and connects all of this data so that it becomes more structured and consumable for our AI services.

The second component is our finance agents. These are the AI automations built on top of the knowledge graph, designed to execute tasks with extremely high accuracy and minimal hallucination.

For example, when analyzing securities agreements, our knowledge graph indexes and understands all of the complexity. These agreements are typically 150 pages long with a huge amount of logic encoded inside them. Our AI service maps everything out automatically, and our finance agents then interpret that information for use in credit investing, due diligence, or other financial processes.

How are you measuring the impact that Gradient is having on the financial workflows of your key customers? Are you most focussed on speed, accuracy or security? 

Within back office use cases, many of these processes generate value for the business by enabling scale and efficiency. Previously, for tasks like suspicious activity reporting or compliance monitoring, companies needed large operational teams to manually review all this information. These were high-skill workers who were difficult to hire, and achieving full coverage was a challenge. What we’ve enabled is scale and efficiency in that space.

On the front office side, it’s less about efficiency and scale and more about augmentation. Investment analysts need to cover more securities and investments with higher accuracy and better decision-making because that’s the direct profit driver for many funds. What we’re doing is providing additional intelligence, delivering new information that helps them make decisions, and helping them mitigate risk. That’s really where we provide value.

How have you navigated AI adoption in finance - one of the most regulated industries in the world? Were there specific hurdles you had to overcome? 

I think there are really two camps for AI adoption today. One is the class of problems where you can fully automate using AI, and the second is where an individual expert still serves as the final backstop. I break it down into those two categories.

For full automation, many of these tasks need to be exceptionally high accuracy and have clear rigor behind their definitions. What we do is something similar to integration tests. Once we deploy an automation, we create a structured set of golden data that we use as an evaluation test on a recurring basis. This provides quantitative metrics to ensure that the service is maintaining 98–99% accuracy on a given task, and it allows us to detect issues as they arise.

For cases where humans remain the final backstop, we provide a copilot experience. The key value here is traceability and auditability. Every time an AI makes a recommendation or an assessment, we provide full visibility into how that conclusion was reached. You can trace the sources used, follow the system’s chain of thought, and understand exactly how the AI derived its answer. All of this is transparent and can be reviewed by the individual expert.

2025 seems to be the year of the AI Agent. How are you thinking about utilizing agents in Gradient’s product and internally at the company, especially given how highly-regulated finance can be? 

I work from the security point up. All of our customers in financial services have very stringent security requirements. At Gradient, we’re SOC 2 Type 2 Compliant and we’ve optimized our product for both cloud agnostic and on-prem deployment from the start.When we deploy into a customer’s environment, the data never leaves their system. That’s how we provide the isolation and guarantees companies need to feel comfortable.

In terms of how we think about the utilization and development of agents, there are really two main approaches. First, we use agents to construct our knowledge graphs. Their role is to understand the underlying essence of the data and infer connections across documents, within documents, and so on, using an agent service.

The second way we develop agents is for specialized functions. For example, we have a compliance agent, a credit agent, and a private equity agent. It’s very difficult to build something that fully replicates a human role, so we hyper-focus each agent’s capabilities on a few core skill sets.

For instance, our credit analyst agent is designed specifically to interpret and understand covenants and how they impact investment risk. Our compliance agent focuses on interpreting regulatory frameworks, policies, and related requirements. By defining very clear tasks, we’re able to make these agents much more powerful, but specialization is essential.

What has been the hardest technical challenge around building Gradient into the platform it is today? 

In constructing these agentic knowledge graphs, one of the key challenges is expanding and scaling them into a more generalized system. The main reason for this is that you have to start blending in traditional information retrieval techniques while ensuring the system remains exceptionally scalable. AI isn’t particularly fast when it comes to understanding and mapping out content today, so we’ve had to build small language models specifically for graph construction tasks to achieve higher speed while maintaining reasonable accuracy. These are some of the general technical hurdles we’ve encountered.

As we expand the universe of information we’re incorporating, we also have to continue investing in reducing latency. Another major challenge is integrating real-time data. It’s not just about indexing or pre-indexing a large corpus of information—like a company’s diligence documents—but also about bringing in real-time market data, proprietary data streams, and other relevant information dynamically. That’s a technical challenge we’ve been actively working on.

Finally, as we move toward higher-order reasoning and more complex agent chains, the risk of AI systems collapsing on themselves increases. Developing strong self-correcting mechanisms to manage these risks is another area of active research for us.

How do you see Gradient evolving over the next 6-12 months? Any specific developments that your customers should be excited about? 

The main focus right now is bringing this into a highly scalable ecosystem—making our product more widely available and expanding the feature set that lowers the barriers for customers and even individuals to leverage our platform. That’s a big priority for us.

The second focus is further increasing and expanding the platform’s capabilities, allowing more use cases to be developed on top of it. We’re constantly working to make the system more powerful and adaptable to a broader range of financial applications.

What was your overall reaction to the DeepSeek moment a few weeks ago, and has that changed anything about your approach to open-source vs. closed-source? 

It’s really interesting because I think there’s quite a bit of misunderstanding about how these models work. For example, after DeepSeek released R1, several of our customers emailed us to verify that we weren’t using DeepSeek under the hood. They were concerned about data being logged by the Chinese government.

Our response was that, first, if you’re privately hosting an open-source model, it’s actually not possible for data to be extracted. And second, it revealed a lot about how they perceive security.

At the end of the day, these companies trust the cloud service providers—Azure, Google, AWS. Anything authorized or verified by these platforms is what they consider secure. I used to advocate for open-source as the space where you have full control and ownership, but over time, working with clients has shown us that trust really lies with these hyperscaler partners.

That has influenced our approach to platform development. We prioritize working within the constraints of the customer’s security requirements. We’ve also designed our platform to be model-agnostic, meaning we can support Azure OpenAI, Gemini, and others. Under the hood, it’s similar to Perplexity in that we add value on top of these foundational models, and our system is fully plug-and-play.

Lastly, how would you describe the culture at Gradient? Are you hiring, and what do you look for in prospective team members joining Gradient? 

We have 26 people on the team today, primarily based in the Bay Area, and we’re increasing our presence in New York. Our team is a hybrid of AI experts and financial services experts, which is exciting because the fusion of these two fields is where real innovation happens.

We’re actively hiring and looking for people interested in using AI to tackle frontier reasoning problems and other complex challenges. We’re specifically hiring full-stack engineers and AI engineers as we continue developing and expanding the service.

Conclusion

Stay up to date on the latest with Gradient, learn more about them here.

Read our past few Deep Dives below:

• Build specialized RAG agents with Contextual AI 🌐

• Remyx - Your AI Production Assistant 💡

• Qdrant's GPU-accelerated vector indexing is here 🔋🌐

• Athena is your AI-powered remote hire 🎛

• Doowii - your AI-first education platform 📖

If you would like us to ‘Deep Dive’ a founder, team or product launch, please reply to this email ([email protected]) or DM us on Twitter or LinkedIn.

Join Slack | All Events | Jobs