AI & Backend Developer with 2+ years of professional experience designing and deploying scalable, production-grade AI systems and backend services. Specialised in building REST APIs and microservice architectures, real-t…
AI & Backend Developer with 2+ years of professional experience designing and deploying scalable, production-grade AI systems and backend services. Specialised in building REST APIs and microservice architectures, real-time audio processing pipelines, LLM integrations, and Retrieval-Augmented Generation systems. Proven track record of owning projects end-to-end, from backend architecture and API development through to cloud deployment and production monitoring, with a strong foundation in Python, Docker, and distributed workflow orchestration.
Engineered high-performance gRPC services and REST APIs for real-time voice processing, achieving sub-50ms end-to-end latency for live audio streams at scale.
Designed and integrated custom processors within the Pipecat microservices pipeline to orchestrate multi-service data flow, enabling the MYNA platform to handle concurrent voice sessions reliably.
Led the training and production deployment of custom TTS models.
Built auxiliary models for emotion recognition and language detection, improving call naturalness scores by ~25%.
Architected an automated call evaluation framework scoring voice-call success through intent analysis and response accuracy.
Optimised LLM prompts to reduce hallucination rates in production dialogues.
Implemented Google Calendar API handlers for automated scheduling workflows.
Integrated Meta Graph API to extend platform reach across Instagram and Messenger.
Contributed to the design and development of a subscription-based billing system supporting recurring SaaS payments, handling financial data flows and transaction logic.
Optimised asynchronous real-time data flow using WebSockets, improving responsiveness and connection stability across concurrent sessions.
Engineered high-performance gRPC services and REST APIs for real-time voice processing, achieving sub-50ms end-to-end latency for live audio streams at scale., Designed and integrated custom processors within the Pipecat microservices pipeline to orchestrate multi-service data flow, enabling the MYNA platform to handle concurrent voice sessions reliably., Led the training and production deployment of custom TTS models; built auxiliary models for emotion recognition and language detection, improving call naturalness scores by ~25%., Architected an automated call evaluation framework scoring voice-call success through intent analysis and response accuracy; optimised LLM prompts to reduce hallucination rates in production dialogues., Implemented Google Calendar API handlers for automated scheduling workflows; integrated Meta Graph API to extend platform reach across Instagram and Messenger., Contributed to the design and development of a subscription-based billing system supporting recurring SaaS payments, handling financial data flows and transaction logic., Optimised asynchronous real-time data flow using WebSockets, improving responsiveness and connection stability across concurrent sessions.
Designed and deployed a scalable enterprise RAG platform using open-source LLMs and Milvus vector database with hybrid semantic search, enabling employees to query 100+ internal documents with high accuracy.
Integrated Nvidia NeMo Guardrails to secure LLM outputs and eliminate prompt injection risks in a regulated production environment, reducing security incidents to zero.
Architected multi-node workflow orchestration using Apache Airflow, Celery, and RabbitMQ, containerised with Docker, improving pipeline reliability, retry handling, and operational resilience across distributed nodes.
Applied quantization techniques (INT8/GGUF) to enable high-throughput LLM inference on CPU-bound infrastructure, reducing inference costs while maintaining response quality.
Designed and deployed a scalable enterprise RAG platform using open-source LLMs and Milvus vector database with hybrid semantic search, enabling the employees to query 100+ internal documents with high accuracy., Integrated Nvidia NeMo Guardrails to secure LLM outputs and eliminate prompt injection risks in a regulated production environment, reducing security incidents to zero., Architected multi-node workflow orchestration using Apache Airflow, Celery, and RabbitMQ, containerised with Docker, improving pipeline reliability, retry handling, and operational resilience across distributed nodes., Applied quantization techniques (INT8/GGUF) to enable high-throughput LLM inference on CPU-bound infrastructure, reducing inference costs while maintaining response quality.
The AI-first Recruitment Operating System that unifies and automates the entire hiring lifecycle for modern enterprises worldwide.
Dr Job FZ LLC. 2026 © All Rights Reserved
