Eagle Alpha’s Alternative Data Conference in New York on February 12, 2025, brought together industry leaders, investors, and data providers for a day of in-depth discussions on the evolving alternative data landscape including AI, consumner insights, and market trends. The conference covered key themes, including the increasing regulatory scrutiny on AI and alternative data usage, insights into hedge fund technology, and the role of data in private equity. A dedicated compliance panel explored the challenges of navigating SEC examinations and the shifting role of regulators, while the wine investment session highlighted how data-driven strategies are reshaping alternative asset investing. A summary of selected discussions is presented below.
Engineering Tomorrow’s Hedge Fund Technology: Avoiding Data Infrastructure Lock-in and Deploying Useful AI Applications
In this wide-ranging first session, our CTO Thomas Combes and Michael Watson, Founder of Hedgineer.io explored how open source data technology is allowing hedge funds to escape vendor lock-in, including how open table formats like Apache Iceberg are causing a convergence between data lakes and data warehouses, and how hybrid compute engines such as DuckDB are allowing data analysis across any environment, local or cloud. They suggested that the value proposition of closed-source platforms like Snowflake might shift primarily to user experience, as open-source alternatives continue to gain traction.
Michael emphasized the importance of first identifying foundational deterministic processes and implementing AI around the interfaces to these processes instead of wholesale replacement of working processes. He also shared an innovative use case where LLMs could monitor portfolio risks by comparing internal research against real-time public information, while also intelligently managing alert fatigue through smart notification systems when research gaps are identified.
The session concluded by stressing the importance of being clear-eyed about accuracy requirements when implementing AI solutions in different parts of a hedge fund’s technology stack, and expressing optimism that the continued development in open-source data technology would bring broad benefits to the hedge fund space.
In Pursuit of Alpha: With Bloomberg Research Data for Systematic Research and AI Workflows
In a time when data is the foundation of innovation, having access to comprehensive, high-quality datasets is essential for driving innovation and informed decision-making. Bloomberg’s Enterprise Data Solutions for Investment Research offer seamless integration of structured financial, alternative, and ESG data, enabling firms to unlock deeper insights at scale.
In this session, Michael Beal, Co-Head of Bloomberg’s Enterprise Data Science Specialists, walked through a variety of machine learning and Generative AI use cases powered by Bloomberg Enterprise’s AI-ready Data. He discussed Bloomberg’s Enterprise Data “Knowledge Graph”, showcasing how research, pricing, corporate actions, and regulatory data intersect to provide deep insights for investors.
One key focus of the session was how structured and unstructured data can enhance financial modeling. Michael discussed the use of embeddings and vector-based approaches, explaining how these techniques help map relationships within financial data, including covariance structures in investment models. Overall, the session reinforced the growing importance of advanced data strategies in quantitative investing, showcasing how Bloomberg’s data solutions support more informed decision-making in an increasingly complex market.
The Opportunities and Nuances of Deploying Unstructured Data in Quant Trading Models
With Professor Francesco Fabozzi, Yale University, Didier Lopez, OpenBB and Dan Joldzic, Alexandria Technology, moderated by Christos Koutsoyannis, Atlas Ridge Capital
The panel spent most of the time discussing the current state of the LLM and NLP landscape and how things are evolving. All the speakers mentioned that it is important to be “disciplined” on your data sources and inputs. While it may seem intuitive to gather vast amounts of information, the speakers emphasized the need to be selective. Simply deploying a massive corpus of data without a clear objective can lead to inefficiencies, noise, and inaccurate insights. Instead, firms should begin by defining their specific needs and curating high-quality, labeled datasets to improve model performance.
One of the key debates centered around the effectiveness of general-purpose LLMs versus smaller, specialized models. The panel largely agreed that open-source LLMs can be valuable, especially when layered with AI agents for specific tasks. However, Dan from Alexandria Technology pushed back on this notion, arguing that smaller, domain-specific models tend to be more accurate and useful for financial applications. General LLMs, he noted, often provide overly broad insights that lack precision. That said, other panelists countered that understanding how to refine prompts and “finesse” answers could make larger LLMs work effectively in finance.
Beyond model selection, the discussion shifted to data ingestion and security. The speakers underscored that integrating multiple data sources requires careful management, especially when working with external NLP or AI providers. A major consensus was that on-premises deployment is the optimal approach to ensure data security. While cloud-based solutions may be more convenient, they come with risks of data leakage and regulatory exposure. Despite the added complexity and cost, an on-prem setup is often necessary—if not essential—for financial institutions handling sensitive data.
Another crucial takeaway was that firms must carefully balance automation with human oversight. While LLMs can enhance decision-making and streamline analysis, they should not operate unchecked. The ability to validate AI-generated insights is key, particularly in finance, where even minor errors can have significant consequences. Having structured processes to refine and validate AI outputs ensures that firms can benefit from LLMs without compromising data integrity or investment accuracy.
Consumer Insights and Product Developments from Leading Vendors
With Jon Ciaio, Facteus, Thomas Grant, Apptopia and Alex Nisenzon, Charm.io
After the introductory vendor overviews, the panel jumped right in to talk about recent consumer trends. As both Nike (NKE) and Lululemon (LULU) are topical at the moment, the vendors focused on the athleisure market. Tom from Apptopia noted weakness for LULU in the US but hedged slightly by saying the year-on-year comps are hard to get a read on. On the other hand, NKE app data trends are quite negative. This reinforces the sales performance of the company in recent quarters. Tom emphasized Nike’s declining popularity among Gen Z consumers, a demographic shift that could pose long-term challenges.
Facteus data supported the comments on NKE with weak spending and weak ASP trends of late. Gen Z spending is also noticeably weak. However, Jon pointed out that LULU experienced a relatively strong holiday season in the US. Charm.io brought an additional layer of insight through TikTok data, revealing that while major brands like Nike and Lululemon have a limited presence on the platform, smaller, digitally savvy athleisure brands are gaining traction. Alex detailed how Charm.io has access to TikTok Shop data across the US, APAC, and select European markets, providing a unique view into emerging retail trends.
On the topic of TikTok itself, Alex said that they see very strong spending trends globally and that the potential ban in the US had no impact on recent sales or over the holiday period. Both Facteus and Apptopia also confirmed the growth of TikTok with very strong trends at the card spend level and app downloads and DAU’s.
Discussion then turned to the potential impact on consumer spending for ETSY, AMZN and other indirect sellers if the US government changes the “de-minimis import exemption” on low imports less than $800. If this exemption gets curtailed it could have a large impact on Temu and Shein imports to the US. This could then benefit sellers like ETSY.
Tom gave some stats on this where 50% of all ETSY users use Temu and 30% use Shein. Facteus said they saw weakness in the holiday period for ETSY in general. As for the impact of any exemption they do not see any GenZ data on ETSY so the impact of Shein and Temu on that demographic is likely diminish. Alex said that TikTok’s generally very low ASP transactions for both the US and globally would show up in their data.
2025 Compliance: What to Expect from the New Administration
With Sergio Pagliery, Schulte Roth & Zabel and Emilie Abate, Iron Road Partners
The compliance panel explored the evolving regulatory landscape, with Emilie highlighting a key shift: it has become significantly more difficult to initiate an investigation compared to previous years. Now, investigations require approval at a much higher level, making the threshold for scrutiny notably more stringent.
The discussion underscored the necessity for robust policies and procedures, not only for the use of alternative data but also for AI-driven processes within funds. There are two critical areas of focus: internal AI applications within the fund and the due diligence required when assessing third-party providers, such as data vendors, for their AI usage.
A crucial point raised was the nature of SEC examinations. These are inherently backward-looking, meaning that while regulatory intensity may seem to be decreasing now, funds will still be assessed based on their compliance frameworks from two or three years ago. This underscores the importance of maintaining stringent policies proactively, rather than reacting to the current regulatory climate.
The panel also explored a potential shift in the SEC’s role. Rather than acting purely as an enforcement body, the SEC could evolve into more of a consulting entity, working alongside funds to ensure best practices in compliance and regulatory adherence.
Alien Intelligence, Myth and Reality
In a comprehensive presentation on AI development, Frederic Siboulet discussed the evolution and future trajectory of large language models, with a particular focus on their generalization capabilities and computational requirements. He explained how newer models like OpenAI’s o3 reasoning model were achieving breakthrough performance on intelligence benchmarks like ARC-AGI, while emphasising that current models appeared to be undertrained, with too many parameters relative to the size of their training data.
Frederic highlighted findings from prominent academic research, such as the Chinchilla paper’s finding that there was an optimal tradeoff where reducing model size and training longer on larger datasets led to better performance, suggesting that data scarcity rather than model architecture might be limiting progress. He detailed analysis of model architectures, training costs, and inference capabilities, noting that a smaller model (by parameter count) trained on more tokens outperformed a larger model trained on fewer tokens, indicating significant inefficiencies in how models were being scaled.
The presentation identified several potential solutions to this undertraining problem, including the use of synthetic data generation and hybrid approaches combining real-world data with high-quality synthetic augmentation. For financial services AI applications, Frederic outlined a risk matrix breaking down large language model capabilities vs. their impacts to clients, front-office, and back-office functions, and noted the current model accuracy is significantly below the threshold expected for existing critical infrastructural systems such as passenger air travel, underscoring the progress yet to be made to AI models before achieving mainstream acceptance.”
Uncorking Alpha: Data-Driven Strategies in Fine Wine Investing
With Tommy Jensen, Wine Capital Fund
The session on wine investment reinforced a principle that is as relevant to markets as it is to fine wine: never let emotions drive investment decisions. A systematic, data-driven approach to wine investing mirrors traditional equity research and investment strategies, with a particular focus on arbitrage opportunities within the wine market.
Tommy emphasized the Wine Capital Fund’s exclusive focus on investment-grade wines, a category distinct from consumer wines commonly found in retail stores. Investment-grade wines, often produced in limited quantities by prestigious vineyards, command prices well beyond the reach of everyday buyers.
Much like fine art or rare collectibles, these wines serve as an alternative asset class, appreciating in value due to scarcity, brand prestige, and historical performance. The discussion highlighted the parallels between wine and traditional financial assets, showcasing how structured investment strategies can unlock value in this niche market.
New to Market & Fresh Features
In this rapid-fire session, 15 data vendors showcased their latest innovations in just three minutes each. From new-to-market datasets to exciting fresh features, attendees gained insights across a spectrum of updates from the alt data industry.
New to Market:
- Crosswalk: Zero-party clickstream data monitoring the digital behavior of 30M global consumers, providing insights into purchases, e-receipts, demographics, behaviors, and more.
- Podchaser: Comprehensive podcast database with metrics for 200M episodes. Provides listening data, sector trends, C-suite insights, and sentiment analysis for better investment decisions.
- Trading Hours: Global financial market holiday and trading schedule data, meticulously researched and delivered in a machine-readable format to support trading strategies and risk management.
Fresh Features:
- Mill Street: MAER ranking model for over 6,000 global stocks, offering customizable insights from publicly available data for transparency and flexibility.
- AnthologyAI: New 2025 Q1 releases covering consumer behavior data in ride-share, food delivery, and streaming, providing deeper insights into evolving trends.
- Babel Street: Updates following Babel Street’s January 2024 acquisition of Vertical Knowledge (VK), a familiar name for many in the industry.
- Facteus: Launch of Onyx, a US retail POS transaction panel capturing $400B in spend across 50K store locations, covering categories like home improvement, apparel, and CPG.
- Gridwise: Gig mobility data to track brand performance, delivery hotspots, and on-demand delivery trends across 90+ tickers for better investment decision-making.
- Jetnet: Introduction of Wingx, a global air freight tracking dashboard with real-time aggregation of flight activity flows by region, country, and route pairings.
- Orbit: AI-powered platform with Chinese data, global filings, sustainability documents, and broker research for enhanced investment and business insights.
- Smart Insider: New Cannibal Score feature added to the share buybacks dataset, measuring how aggressive a company’s buybacks are relative to peers for improved buyside decision-making and investment outperformance
- Yukka Labs: AI-driven event detection engine processing real-time data streams for actionable insights on Risk, ESG, and sales, tailored for investment managers.
- G2: Introducing B2B Spend and Contract dataset offering visibility into software licensing, seats, renewals, and more, empowering better software investment decisions.
- 3Ai: Introducing several updates, including Daily Alpha Forecasts and Business Cycle AI which offers 30% more explained variation than human classification. Next year, they’ll update signal forecast timeframes to 1w, 1m, 3m, 6m, 9m.
- Flywheel: Launch of eCommerce 2.0, a global pricing and promotion data suite developed through extensive scraping, mapping, tagging, and transformation of hundreds of websites.
