Substance Use Survey Anomaly Detector with AutoGen and Featherless AI
Inspired by recent NIH findings on elevated cannabis and nicotine usage trends among US adults, this challenge focuses on constructing an agentic public health surveillance system. Using AutoGen multi-agent conversation framework paired with low-latency model inference provided by Featherless AI, participants will build collaborative agents (Survey Parser, Anomaly Detector, Policy Analyst) that ingest high-volume survey feeds to detect emerging regional consumption anomalies.
What you are building
The core problem, expected build, and operating context for this challenge.
Deploy collaborative AutoGen agents powered by Featherless AI to monitor public health survey anomalies.
How work is evaluated
Evaluates multi-agent conversation accuracy in detecting statistical anomalies in survey time-series.
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.
detected_anomaly_present
Ensure spike on 2025-02 is flagged correctly
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
anomaly_precision
Precision of detected statistical anomalies against synthetic ground truth • target: 0.9 • 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
Configure multi-agent multi-turn communication using Microsoft AutoGen
Connect Featherless AI inference runtime endpoints for open-weights LLMs
Perform automated temporal time-series anomaly detection on epidemiologic data
Generate actionable public health advisory reports based on automated findings
Reference links and supporting material
Synthetic time-series dataset of regional adult survey responses regarding monthly cannabis and nicotine consumption frequencies.
Challenge setup
Evaluates multi-agent conversation accuracy in detecting statistical anomalies in survey time-series.
Challenge input
JSON array of monthly survey metrics by region
AutoGen
Mandatory agent framework
Featherless AI
Assigned model provider
Evaluated output
JSON object with anomalies_detected (list) and recommendation_summary (string)
- Ensure spike on 2025-02 is flagged correctly
- Precision of detected statistical anomalies against synthetic ground truth • target: 0.9 • range: 0-1
- Benchmark: SurveillanceBench
- Anomaly Precision target: 0.9
- 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
- Featherless AI · Required
- AutoGen · Optional
- Framer · Optional
Evaluation contract
- detected_anomaly_present · Weight 1
- anomaly_precision · Weight 1
Metrics
- anomaly_precision · Target 0.9
Benchmarks
- SurveillanceBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON array of monthly survey metrics by region
Output: JSON object with anomalies_detected (list) and recommendation_summary (string)
Execution harness: Python
Evaluation tools
- AutoGen · Agent Framework
- Featherless AI · Inference Infrastructure
Recipe freeze marker
No recipe freeze time is recorded.