Why moving models into production is an infrastructure, latency, and data hygiene challenge—and what I will be documenting on this publication.
Building production AI systems: Speech AI, Agentic Workflows & Resilient RAG
Production-focused AI Engineer specializing in Speech AI (ASR), Agentic Architectures, and low-latency retrieval systems. I design, deploy, and benchmark AI systems with an emphasis on production constraints, latency minimization, and data pipeline hygiene. Core focus areas & technical proof: Speech AI & Dialect Adaptation: Fine-tuning Whisper-small architectures for conversational speech, raw audio signal filtering (~53 GB to 7.7 GB pristine subsets), and Hugging Face deployment. Agentic Orchestration & Async State Machines: Architecting zero-disk I/O binary processing pipelines and resilient multimodal content engines with n8n and foundation model APIs. Production RAG Pipelines: Engineering low-latency inference architectures with dynamic 3-hour vector synchronization and decoupled post-session analytics. Core Stack: Python, PyTorch, Hugging Face (Transformers / Datasets), Gemini APIs, OpenRouter, n8n, Docker, Vector Databases.
Markdown supported · ⌘↵ to post
No responses yet.