I worked on enterprise analytics and AI at SAP. Before that, I built financial models and software to measure portfolio risk, value bonds, and analyze trading performance. I'm currently building applications for working with AI, investigating operational problems, and making decisions from data.
Coding agents can change an entire codebase in minutes. Understanding and reviewing their work is much harder.
League of Agents lets you run multiple coding agents in the same repository, explore your project as a zoomable map, direct agents to specific files or lines, and watch their changes as they happen. Review diffs, run tests, and keep or undo each run.
Many businesses still rely on software with no API, where everyday tasks require someone to navigate screens and enter information manually.
Eagleview uses an AI agent to discover how to perform a task, converts the successful workflow into a program that runs without the model. Each step verifies the application's state, with automatic recovery, human takeover, and approval for irreversible actions.
The quality of an AI's response often depends on how clearly you communicate what you need.
Averroes adds a separate AI coach to a chat application. It reviews your conversation, identifies missing context or unclear instructions, and suggests better prompts. The 0 → 1 mode helps brainstorm rough ideas into clear instructions before you start something bigger.
When a supplier's invoice doesn't match what a company ordered or received, an analyst has to investigate before payment can proceed.
This application compares invoices, purchase orders, and goods receipts across 14 discrepancy types. It identifies the cause, presents supporting evidence, and recommends a resolution. Financial rules determine the outcome, with human approval required for corrections.
Companies can meet staffing targets by hiring externally, moving existing employees, or training people for new roles. Finding the right combination requires understanding available skills, costs, and deadlines.
Given a staffing target and budget, this application identifies qualified employees, training opportunities, and remaining hiring gaps. It compares costs and timelines, checks the evidence behind employee skills, and produces a plan using fixed, auditable rules.
- mbs-val: Mortgage-backed security valuation
- dtd: Distance-to-default estimation using market-value and volatility-based methods
- us-delinquency-forecast: Forecasting U.S. delinquency rates from economic data
- strat-backtest: MATLAB application for backtesting MACD and RSI strategies with portfolio P&L