We're thrilled to announce that Joseph Kim has joined us as our Head of Applied AI. With his background in AI and search – from MIT to NASA to Google – Joseph brings invaluable experience to help advance Rogo's mission of transforming finance with AI.

From Space Robots to Search: A Journey in Technical Innovation

MIT & NASA

Joseph began his career as a researcher at the MIT Media Lab, working on programmable metamaterials — an interdisciplinary field combining mechanical engineering and computer science. After graduation, he spent four years at NASA as a research scientist, developing systems for aeronautics and robotically assembled space structures. This work involved early applications of deep reinforcement learning, technology that would later influence developments like AlphaGo.

Retail & Industry-Specific AI at Google

In 2017 and 2018, the transformer and BERT papers were released. "Everyone in the ML community had our 'ChatGPT moment' then," Joseph says. Following this development, Joseph joined Google to focus on artificial intelligence. He started in retail search, where he gained insights into industry-specific AI solutions.

"Given the critical impact of online product search on a retailer's revenue and profitability, the lack of technical expertise and investment in this area was shocking," Joseph explains. "Companies relied on thousands of brittle and hand-tuned rules, instead of leveraging the power of big data and machine learning to automatically optimize revenue, and most retailers put very little emphasis on hiring top engineers or developing truly scalable, impactful search solutions. Even with relatively simple ML techniques, Google could outperform in-house solutions by up to 50%. It opened my eyes to how underserved high-stakes industries can be when it comes to applying cutting-edge technology."

Web Search and Gemini at Google

After growing Retail Search from a prototype into one of Cloud's most profitable businesses, Joseph moved to Google's core ranking and ML team for web search. His initial focus was launching BERT to Search's billions of users. "Introducing transformer models into Google Search led to the biggest step-function improvement to search quality ever, and bringing those semantic understanding improvements to Search's billions of users worldwide was a deeply rewarding and humbling experience," Joseph recalls.

Following ChatGPT's launch in late 2022, Joseph joined the Gemini team at Google. His research focused on reinforcement learning with human and machine feedback for improving generative language models.

"The first major wave of AI capabilities was unlocked by the invention of the transformer, enabling super-human semantic understanding at scale. The subsequent release of RLHF and ChatGPT in 2022 further revolutionized the field, enabling sophisticated instruction-following that spawned a host of new applications and products. We are now on the cusp of the third wave: agentic AI. These systems will move beyond simple chat responses to actively participate in complex tasks, working alongside humans in a symbiotic relationship. This shift promises to reshape workflows across numerous sectors, with finance poised to be at the forefront of this transformation."

Why Finance Needs a Search Revolution

Joseph's interest in Rogo stems from his observations about technology in finance. Despite managing trillions of dollars, many financial firms continue to rely on manual processes and fragmented tools.

"I saw the same pattern in finance that I witnessed years ago in retail search," Joseph notes. This insight became clear during the pandemic when a hedge fund offered $20,000 a month for software to track stock discussions on WallStreetBets. "That's a 3-hour project for an engineer that a hedge fund would pay $20,000 a month for, likely because they lack the expertise to even know if it was a 3-hour or 3-month project."

Building the Future of Financial Intelligence at Rogo

At Rogo, Joseph will lead our AI development, focusing on innovations including:

• Multi-step reinforcement learning with human and machine feedback

• Custom-built hybrid retrieval optimized for LLM prompt packing (RAG++)

• Monte Carlo Tree Search for scaling test-time compute

"Finance has always been about rigor and precision," Joseph adds. "At Rogo, we have the opportunity to bring modern AI capabilities to the table."

Join Our Team

If you're interested in applying cutting-edge AI to complex financial problems, we're hiring.

"Within Google Search and Gemini, we tackled some fascinating technical problems. However, the sheer size of Google sometimes meant that there weren't enough GPUs in the world to build our exciting research prototypes into real products. At Rogo, the economics around cost per query are enormously different – and I'm excited to explore a novel solution space and to play a deep role in shaping the future of financial work."