Accelerating Investment Decisions with Specialized Claude Code | Databricks Data & AI Summit 2026
In this session from Databricks Data + AI Summit, Maciej Szpakowski (Co-founder, Prophecy) and Shehzad Nabi (CTO, Waterfall Asset Management) walk through how Waterfall partnered with Prophecy to build a specialized AI platform for structured finance, replacing hours of manual "tape cracking" with a governed, AI-native workflow that runs on Databricks. In private credit and structured finance, deal data arrives in any format imaginable — spreadsheets, CSVs, PDFs — with no standardization. Portfolio managers and traders can't act until that data is clean, and the clock is always ticking. This session shows how Prophecy's AI changes that equation. You'll see a live demo covering the full workflow: ingesting a messy loan tape, AI-generated column mappings with confidence scoring, ad hoc analysis and data cuts, unstructured document processing, and automated portfolio surveillance — all built on Databricks with code that's fully inspectable and stored in Git. Key topics covered: • The "tape cracking" problem in structured finance and why existing tools fall short • Why Claude on its own wasn't enough • Confidence scoring: how Prophecy tells you which AI outputs to trust and which to review • Live demo: harmonizing a messy loan tape in minutes vs. hours • AI-powered stratifications, data cuts, and ad hoc investment analysis • Unstructured data processing: extracting insights from prospectus documents • Setting up automated portfolio surveillance with self-healing pipelines • Real results: review burden dropping from 145 columns to 4 by the fifth tape • Remaining challenges: context management with large-scale financial data 🔗 Learn more about Prophecy for structured finance: https://www.prophecy.ai/structured-finance 🔗 Waterfall Asset Management: https://www.waterfallam.com Timestamps 00:00 - Introduction to Prophecy’s role in accelerating structured finance investment decisions 02:06 - Challenges in private credit data access and normalization 04:01 - Manual tape processing workflow and its limitations 06:00 - Building a custom AI-powered platform with Prophecy for data workflows 07:26 - AI's capabilities in tape cracking, mapping confidence, and explanation 09:03 - Analyzing clean data and AI-backed portfolio insights 10:00 - Data surveillance and automatic anomaly detection 11:24 - Collaboration improved with a shared platform now replacing spreadsheets 13:18 - Technical architecture: data studio, AI pipelines, and integration with Databricks 14:23 - Demonstrating automation: onboard new tapes, minimize reviews 16:48 - Challenges with large data and LLM integration in finance workflows 18:20 - Live demo: messy data onboarding, harmonization, and analysis in Prophecy 29:27 - Extracting insights from unstructured PDFs and setting up surveillance pipelines 32:18 - Wrap-up: the value of human-in-the-loop AI in structured finance