Untitled

package com.insuretech.claims.fraud;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.cache.annotation.Cacheable;
import org.springframework.stereotype.Service;
import org.springframework.web.reactive.function.client.WebClient;
import reactor.core.publisher.Mono;

import java.util.Map;

@Service
public class FraudDetectionService {

    @Autowired
    private WebClient webClient; // Configured for SageMaker endpoint

    @Cacheable(value = "fraudScores", key = "#claimId") // Redis-backed cache for optimization
    public Mono<Double> scoreClaimFraud(String claimId, Map<String, Object> claimData) {
        // Enrich with features: e.g., claim amount, location anomaly, telematics velocity
        Map<String, Object> features = enrichFeatures(claimData);
        
        // Reactive call to ML model (non-blocking for high throughput)
        return webClient.post()
                .uri("/predict") // SageMaker endpoint
                .bodyValue(features)
                .retrieve()
                .bodyToMono(Double.class)
                .doOnNext(score -> {
                    if (score > 0.8) { // Threshold for high-risk
                        // Trigger alert via SNS (async)
                        publishAlert(claimId, score);
                    }
                })
                .onErrorReturn(0.0); // Fallback for resilience
    }

    private Map<String, Object> enrichFeatures(Map<String, Object> claimData) {
        // Simulate anomaly detection: e.g., velocity check
        Double amount = (Double) claimData.get("amount");
        String location = (String) claimData.get("location");
        double anomalyScore = (amount > 10000 && "high-risk-area".equals(location)) ? 0.5 : 0.0;
        claimData.put("anomalyBase", anomalyScore);
        return claimData;
    }

    private void publishAlert(String claimId, Double score) {
        // Async publish to SNS for notifications (decoupled)
        // Implementation omitted for brevity
    }
}