Tech People in the Know: Delfi’s Joseph Ahn

In what is a recurring feature, Finopotamus profiles interesting and intriguing tech professionals who are positively impacting the credit union industry. For this issue, we visited with Delfi Co-Founder and Chief Strategy Officer Joseph Ahn. The New York City-based fintech company bills itself as “revolutionizing financial risk management by bringing artificial intelligence (AI)-enabled algorithmic derivative hedging strategies to enterprises of all sizes.”

By W.B. King

As a data scientist, Joseph Ahn has long held an interest in emerging AI and machine learning (ML) technologies. While finishing his Ph.D. at Harvard Business School, he was consumed with analyzing patents and technology transfers using early-stage AI language models.  

Joseph Ahn

“The power of data at volume is clear. We were watching the victory of Google’s AlphaGo AI over the world. ‘Go’ was historically considered almost ‘unsolvable’ due to the sheer size of the possible playing space,” Ahn told Finopotamus. “We realized then that AI was mature enough to tackle our core problem of balancing portfolios for optimal risk management.”

Prior to cofounding Delfi in 2022, Ahn worked for Deloitte as a manager, as an adjunct professor at Cornell University and as a lead data scientist and senior economist for economic policy and the U.S. Office of Management and Budget.

Cannonballs to Kill Mosquitos

Over the last 10-plus years, Ahn has noticed a few important changes in the fintech space, including a broader acceptance and trust in data science and AI.

“When I started, many models had to be justified from the ground up. Today, organizations not only understand the value of AI—they actively seek to integrate it into strategic decision-making,” he shared. “At the same time, there’s greater awareness of the critical importance of high-quality data. Tech teams are placing more emphasis on how data is collected, labeled, and preserved—recognizing that flawed inputs lead to flawed outputs.”

With advancements come challenges, he noted, including an overreliance on AI, “particularly when it’s applied without context or proper validation. AI models are only as useful as the data they’re trained on.”

Another challenge, he said, is AI's inability to define patterns that have occurred in the past but don’t appear in data edge cases. “Without proper controls, AI can fail in spectacular ways, like promising products or services that don’t exist,” he continued. “AI is also notoriously bad with numbers and should be backed with an outside ‘engine’ or ‘calculator’ to support AI’s interpretive abilities.”

The financial services industry also has a growing need to match the complexity of tools to the complexity of the problem, he said. “In many cases, proven models like XGBoost or decision trees are still the most effective solution. We often see a temptation to apply large generative models where simpler, well-understood techniques would be more appropriate,” Ahn stated. “It’s important to avoid using a ‘cannonball to kill a mosquito.’”

From a management perspective, he said there has been a shift toward cross-functional collaboration and faster iteration. “Teams are more integrated, and tech leaders are expected to align more directly with product and business outcomes than in the past,” he noted.

Trust and Verify

With five tech employees, including two engineers, on staff, Ahn said his team is always open to innovation, while rigorously testing every new solution before adopting it. Evaluation starts with clearly defined success criteria and a set testing period, which allows the team to measure outcomes objectively rather than by “hype or intuition,” he shared.

“We also believe it is essential to invest in the validation process. That means providing access to the right data, securing buy-in from internal stakeholders, and, when properly vetted and justified, committing controlled capital for pilot projects,” Ahn told Finopotamus. “Ultimately, we aim to strike a balance—be open to cutting-edge solutions but disciplined in proving their value before scaling.”

To stay current with tech trends, Ahn also regularly attends industry conferences, which he said are a vital part of staying current in “such a fast-moving space.” These events, he noted, also help the company find potential partners.

“These can include infrastructure providers or fintech innovators who can support our long-term strategy,” he shared. “Recent standout conferences include ‘Acquire or Be Acquired,’ where takeaways included increased emphasis on explainable AI, and much discussion on this new era of market volatility and potential solutions for dealing with that.”

Maintaining Trust and Transparency

Among industry trends on Ahn’s radar is the growing use of AI for data extraction and cleaning, which enables the development of leaner, more efficient ML models.

“This is critical in regulated or resource-constrained environments like credit unions. There’s a larger trend where AI is being used not as a monolithic ‘black box’ but as part of interpretable pipelines broke into modular ‘chunks’ that accomplish specific tasks,” he said. “This makes it easier to audit, refine, and scale individual steps, while maintaining trust and transparency.”

Ahn offered an example: “In a fraud claims processing pipeline, one AI module might clean and normalize data, another might detect anomalies or flag incomplete applications, and a third might use a lightweight machine learning model to assess risk scores,” he said. “Each component can be independently tested, tuned, and explained to compliance teams or boards.”

The application of these pipelines to portfolio risk analysis and operational decisioning areas is especially interesting to Ahn’s team. “Overall, the shift is toward practical AI—solutions that are smaller, faster, and easier to integrate into legacy infrastructure, rather than headline-grabbing general AI.’”

Credit Unions: Co-Investors in New Technologies

What impresses Ahn most about the credit union industry is its inherent focus on the member experience. To this end, he said credit unions drive technological decisions that prioritize member outcomes rather than operational efficiency.

“Credit unions are often more agile than large banks, with fewer legacy constraints, allowing for faster adoption of modern, modular fintech solutions. They have a culture of collaboration not often seen in traditional financial institutions,” he continued. “This is reflected in their participation in CUSOs and shared innovation networks, making them more open to co-developing or co-investing in new technology.”

Over the last three to five years, Ahn said he has seen an uptick in progressive relationships between credit unions and fintechs. “COVID-19 was a major driver; the sudden need for digital service delivery pushed many credit unions to reevaluate their technology stacks. Conversations that once took months now happen in weeks. We recently introduced our newest product, DELFIswaps, which is a form of more lightweight, more convenient swaps,” he said, noting that there is no International Swaps and Derivatives Association (ISDA) requirements or need for hedge accounting.

“We had several conversations with a credit union group that expressed great interest in these DELFIswaps for a variety of use cases, particularly for facilitating matching loan terms to preferences (i.e. maturity and variable vs. fixed),” he told Finopotamus. “The emphasis [for credit unions] is typically on community trust and transparency, which influences how they evaluate and integrate tech—favoring tools that are explainable and measurable in terms of impact on members.”