ML Engineer — Chugani

Overview

Built a real-time liveliness detection service at Chugani using MTCNN and FaceNet, converting research prototypes into modular, production-grade MLOps pipelines. The system became a core component of the identity verification platform, handling high-volume, low-latency inference with continuous monitoring and retraining.

Key Responsibilities

• End-to-End ML Architecture: Designed, developed, and deployed a production-grade liveliness detection pipeline using advanced computer vision techniques, processing over 10,000 daily transactions with 99.2 percent accuracy.
• MLOps Pipeline Engineering: Transformed experimental prototypes into automated, scalable MLOps workflows, reducing deployment time from two weeks to four hours while maintaining 95 percent system uptime.
• Cloud Infrastructure & Monitoring: Built comprehensive AWS-based monitoring and retraining systems with automated performance tracking, anomaly detection, and alerting for proactive issue resolution.
• Performance Optimization: Achieved a 60 percent reduction in inference latency through model optimization techniques including pruning, quantization, and hardware-aware acceleration.
• Enterprise Integration: Implemented robust API integrations and structured logging frameworks enabling seamless communication between ML services, security systems, and compliance platforms.
• Cross-Functional Collaboration: Worked closely with security, compliance, and product teams to align ML delivery with regulatory requirements while meeting aggressive performance and reliability targets.

Technical Achievements

• Accuracy: Built an anti-spoofing pipeline achieving 99.2 percent accuracy in production.
• Latency: Reduced inference latency by 60 percent through model optimization and system tuning.
• Reliability: Established automated retraining and monitoring pipelines with 95 percent uptime.
• Observability: Implemented real-time monitoring with automated alerts and dashboards.
• Scale: Deployed a scalable architecture capable of handling more than 10,000 requests per minute.

Impact

Delivered a robust liveliness detection system that became a critical component of Chugani’s identity verification platform. The solution enabled secure, seamless user authentication while significantly improving system reliability, latency, and operational scalability.

Outcome

This engagement established a strong foundation for production-grade computer vision systems at Chugani, demonstrating how disciplined MLOps practices can transform experimental ML models into reliable enterprise infrastructure.