I’m Sai Teja Reddy, a Machine Learning Engineer focused on Speech AI and LLM fine-tuning, building practical systems that reliably deliver low-latency, high-quality results. I work across model training and evaluation, with hands-on experience using PyTorch and Hugging Face Transformers, and I’ve fine-tuned Gemma variants with LoRA/PEFT methods while debugging hallucinations, persona consistency, latency, and failure cases. On the engineering side, I design GPU-aware inference pipelines and distributed ML infrastructure using Kubernetes, Redis, Kafka, and dynamic batching to improve throughput and robustness during rollouts. I’ve also built end-to-end Voice AI and agentic workflows that combine ASR/TTS, retrieval-augmented context, and production-grade reliability checks—drawing from experience with Whisper, FasterWhisper, StyleTTS2, and Qwen-Audio.…I’m Sai Teja Reddy, a Machine Learning Engineer focused on Speech AI and LLM fine-tuning, building practical systems that reliably deliver low-latency, high-quality results. I work across model training and evaluation, with hands-on experience using PyTorch and Hugging Face Transformers, and I’ve fine-tuned Gemma variants with LoRA/PEFT methods while debugging hallucinations, persona consistency, latency, and failure cases. On the engineering side, I design GPU-aware inference pipelines and distributed ML infrastructure using Kubernetes, Redis, Kafka, and dynamic batching to improve throughput and robustness during rollouts. I’ve also built end-to-end Voice AI and agentic workflows that combine ASR/TTS, retrieval-augmented context, and production-grade reliability checks—drawing from experience with Whisper, FasterWhisper, StyleTTS2, and Qwen-Audio.WWWW…