Skip to content
Sharif Amlani, PhD

Three days of credit-risk work.
Three minutes.

I led generative AI product development for Moody's Commercial Mortgage Metrics platform and shipped three AI agents that took manual credit-risk work off analysts' desks.

Role
Assistant Director of Research
Organization
Moody's Analytics, Commercial Mortgage Metrics (CMM)
Years
2023 – 2025 · San Francisco
Stack
Python · R · API integrations · AI agents

The outcome

Same work.
A sliver of the time.

3 daysManual analyst workflow

The full bar is the three-day manual workflow; the sliver is the three-minute agent run, shown at the minimum visible width.

3 min
Per credit-risk workload
Down from 3 days
20+
Enterprise clients
Using the products
$1T+
In client valuations
Combined, across those clients
3
AI agents shipped
Run CMM · Reporting · Red Flag

01 — The problem

Credit risk ran
on analyst hours.

Moody's Commercial Mortgage Metrics platform supports commercial real estate credit-risk analysis for more than 20 enterprise clients whose valuations exceed $1 trillion. Processing the commercial mortgage data and reporting on it was manual, and a single workload took three days.

02 — What I built

Three agents.
One workflow.

  • Led generative AI product development for the Commercial Mortgage Metrics platform.
  • Developed and deployed all three agents to automate credit-risk workflows.
  • Built the API integrations and standardized the reporting process, in Python and R.
  • Turned quantitative methods into production-grade enterprise data and modeling products.
Agent 01

Run CMM Agent

Runs the CMM credit-risk model, end to end, without an analyst at the keyboard.

Agent 02

CMM Reporting Agent

Analyzes the model results with data analysis and writes them up as standardized reports.

Agent 03

Red Flag Agent

QAs the documents: reads them and flags what an analyst needs to look at.

03 — How it works

Data in.
Reports out.

Fig. 01 — How the agents fit together
Simplified
  1. 01 · IngestAPI integrations bring commercial mortgage data into one standardized process, in Python and R.
  2. 02 · CheckThe Red Flag Agent QAs the documents and flags anything an analyst should see.
  3. 03 · RunThe Run CMM Agent runs the credit-risk model.
  4. 04 · AnalyzeThe CMM Reporting Agent analyzes the results and writes the standardized report.

04 — How we knew it worked

Measured,
not guessed.

An agent that is fast and wrong is a liability. These are the results on the record.

Speed3 days → 3 minOne credit-risk workload: three days as a manual analyst workflow, three minutes with the agents.
Adoption20+ clientsProduction enterprise data and modeling products, used by clients whose valuations exceed $1 trillion.

05 — In their words

Rated exceptional by his manager and colleagues for performance and contributions.

Moody's Analytics · Performance review · 2023 – 2025