Build Distressed Office Asset Screener using Claude Agents SDK & Baseten Engine
With commercial real estate default notices piling up in primary markets like New York, fund managers require real-time automated asset screening agents. This challenge requires building a distressed asset identification agent using Claude Agents SDK deployed on Baseten's low-latency model inference runtime. The agent processes tenant occupancy logs, debt service calendars, and capital market yield curves to calculate a Distressed Asset Score (DAS). It triggers automated workout recommendations, including loan modification terms or mezzanine debt restructuring strategies.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an automated distressed office asset screening agent with Claude Agents SDK on Baseten high-performance infrastructure.
How work is evaluated
Evaluates calculation precision of Distressed Asset Score (DAS) and validity of proposed debt workouts under high vacancy scenarios.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Score Bounds Validation
Ensures distressed asset score falls within expected range (80 - 95)
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Workout Logic Alignment
Alignment of proposed debt workout against industry workout standards • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement Claude Agents SDK with extended thinking enabled for multi-step financial risk reasoning
Deploy custom sub-models and inference tools on Baseten model runtime
Calculate dynamic Distressed Asset Scores (DAS) based on rent roll velocity and debt roll maturity
Generate automated debt workout memos for CRE asset management committees
Reference links and supporting material
Dataset of 80 NYC office property rent rolls, debt schedules, and vacancy projections.
Challenge setup
Evaluates calculation precision of Distressed Asset Score (DAS) and validity of proposed debt workouts under high vacancy scenarios.
Challenge input
JSON with occupancy_pct, upcoming_lease_expirations_pct, debt_interest_rate, debt_maturity_months
Claude Agents SDK
Required framework for multi-step reasoning.
Baseten
Required tool for low-latency custom inference.
Evaluated output
JSON with distressed_asset_score, risk_tier, recommended_workout, target_loan_modification
- Ensures distressed asset score falls within expected range (80 - 95)
- Alignment of proposed debt workout against industry workout standards • target: 0.95 • range: 0-1
- Benchmark: CREDistressBench
- Workout Logic Alignment target: 0.95
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Baseten · Required
- Kore.ai · Optional
- All Hands AI · Optional
Evaluation contract
- Score Bounds Validation · Weight 1
- Workout Logic Alignment · Weight 1
Metrics
- Workout Logic Alignment · Target 0.95
Benchmarks
- CREDistressBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with occupancy_pct, upcoming_lease_expirations_pct, debt_interest_rate, debt_maturity_months
Output: JSON with distressed_asset_score, risk_tier, recommended_workout, target_loan_modification
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
- Baseten · Inference & Model Runtime
Recipe freeze marker
No recipe freeze time is recorded.