Muhammed Junaid K

Ai Engineer @Zoftcares Inc

Kerala, IN

Architected and deployed cloud-native AI systems for real-time video intelligence, multimodal analytics, and ML-driven automation.• Designed streaming pipelines using AWS IVS, Kinesis Video Streams, Lambda, and SQS for high-throughput, low-latency processing.• Built and optimized computer vision pipelines for real-time detection, tracking, and scene understanding using YOLO, PyTorch, and OpenCV.• Improved model efficiency via pruning, quantization, and CUDA optimizations for faster inference.• Developed scalable Python microservices for real-time inference, async processing, and ML orchestration.• Integrated LLM APIs with prompt engineering, caching, RAG, and dynamic model selection to reduce cost and latency.• Engineered AI agent systems using LiveKit with tool execution, enabling workflows, DB operations, and API integrations.• Implemented SIP trunking for voice AI, enabling real-time telephony-based human-AI interaction.• Designed call flows with intent detection, routing, and escalation to human operators.• Delivered integrated video + voice AI systems for alerts, analytics, and multimodal intelligence.• Built resilient, fault-tolerant, message-driven architectures in collaboration with cross-functional teams.Tech Stack & Tools:AWS IVS | Kinesis | Lambda | SQS | Bedrock | LiveKit | SIP | Python | OpenCV | YOLO | PyTorch | TensorFlow | LLM APIs | RAG | CUDA | Microservices | REST APIs | Stream ProcessingKey Focus Areas:#AWS #Kinesis #IVS #Lambda #SQS #Bedrock #LiveKit #VoiceAI #AgentOrchestration #SIP #RealTimeAI #Microservices #LLMIntegration #ComputerVision #CloudAI #VideoIntelligenc