I'm Jeevan Prakash Madabattula — I build production-grade computer vision, voice AI pipelines, and scalable video infrastructure. 3+ years shipping real systems at scale.
End-to-end AI, ML, and infrastructure solutions — ready to deploy.
Object detection, face recognition, ANPR, safety monitoring with YOLO, RetinaFace, TrOCR — deployed at production scale with 95%+ accuracy.
End-to-end multilingual voice agents with STT, LLM reasoning, and TTS. IVR, PSTN/SIP integration via LiveKit, Twilio, and Telnyx.
Scalable VMS supporting thousands of camera streams with MediaMTX, FFmpeg, RTSP/HLS/WebRTC and real-time Grafana dashboards.
TensorRT, ONNX, and Triton Inference Server workflows. Sub-100ms latency and up to 10× speedup on NVIDIA, Intel, and Qualcomm hardware.
Docker, Kubernetes, AWS (EC2/S3/SES), CI/CD, Prometheus/Grafana monitoring, and automated alerting for production AI workloads.
FastAPI/Flask with WebSocket support, PostgreSQL/MongoDB/Redis, and real-time data pipelines — production-hardened and containerized.
Full-stack web apps built for performance and scale — responsive UIs, RESTful APIs, authentication, and cloud deployment. From landing pages to complex dashboards, delivered production-ready.
Cross-platform mobile apps for Android and iOS using React Native. Smooth UX, offline support, push notifications, and seamless API integration — optimized for real-world usage at scale.
End-to-end product development from MVP to production — multi-tenant architecture, subscription billing, role-based access, and auto-scaling infrastructure. Built to grow with your business.
Real systems built and running in production environments.
Distributed VMS supporting 10,000+ cameras across 50+ centers. Real-time Grafana dashboards, AWS SES alerting, and a Bash pipeline that processed 70,000+ video files.
Multilingual voice agent with Indic Conformer STT, Qwen 2.5 LLM, and Svara TTS on LiveKit Cloud. Customer support, sales, and helpdesk agents with auto-recovery.
YOLOv5 + TrOCR pipeline improving accuracy from 30% → 95%. Multi-line plate handling, deployed on NVIDIA Triton for high-throughput inference.
Real-time PPE violation detection with YOLOv5. Automated WhatsApp Business API and email alerts. Containerized for multi-site deployments.
99%+ accuracy face recognition using RetinaFace + InsightFace. Kubernetes deployment with autoscaling and ONNX-optimized Triton inference.
PyTorch/TensorFlow → TensorRT conversion with 5–10× speedup. Sub-100ms latency for complex CV models in production using CUDA optimization.
Technologies I use daily across AI, infrastructure, and backend systems.
Professional journey building AI systems that scale.
Available for freelance projects in AI/ML, computer vision, voice AI, and DevOps. Let's build something great.
Reach out through any of the channels below — I typically respond within 24 hours.