Open to opportunities

Kaushik User

@kaushikuser3

Message

AI/ML engineer building low-latency multilingual voice AI and multi-agent orchestration for production.

India

Message

What I'm looking for

I’m looking to build high-reliability, low-latency voice and agent systems—taking ideas from research to deployment in real-time pipelines, collaborating across teams, and owning model performance, validation, and rollout.

I’m an AI/ML engineer specializing in real-time voice systems and multi-agent orchestration, with hands-on production experience shipping multilingual speech pipelines across 5+ Indian languages.

At Worktual, I engineered a multi-agent CRM orchestration system that automated customer interaction workflows across 3+ pipeline stages—reducing manual intervention by 60% and increasing response throughput by 2.3×.

I validated a multilingual Phone AI Agent with fine-tuned STT/TTS pipelines achieving sub-1.5s end-to-end latency and an 18% accuracy gain over baseline Whisper using sematic VAD, while actively R&D’ing Speech-to-Speech architectures using Mimi codec + LLM inner monologue design on AMD MI300X GPU clusters.

I bring a research-to-deployment mindset for high-reliability, low-latency environments, reinforced by projects like RAG-based CivicBot (document lookup from minutes to ~2s) and my AIDA 2025 work on a hybrid confidence model for mental health detection.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer (AI/ML)

Current

Worktual

Feb 2026 - Present (2 months)

Engineered a multi-agent CRM orchestration system automating customer interaction workflows across 3+ stages, reducing manual intervention by 60% and increasing response throughput by 2.3×. Built and validated a multilingual Phone AI Agent (Tamil, Telugu, Malayalam, Hindi, Kannada) with fine-tuned STT/TTS pipelines achieving sub-1.5s end-to-end latency and an 18% accuracy gain over baseline Whispe

Orchestration STT TTS Fine Tuning Whisper Mimi Codec Speech To Speech (S2S) Systems Sub 1.5s Latency Optimization AMD MI300X GPU Optimization

ML Engineer Intern

Object Automation

Jun 2022 - Aug 2022 (2 months)

Developed and deployed ML models using PyTorch and TensorFlow, improving prediction accuracy by up to 22% across business use cases. Optimized inference pipelines to reduce model latency by 30% while maintaining production performance.

PyTorch TensorFlow