Generative AI Systems
End-to-end orchestration of STT, RAG, and TTS workflows for real-time conversational assistants.
AI Research and Systems Engineering
Shiv Kiran Bagathi
Exploring AI Systems, Stories, and Art.
I design and ship robust software systems for generative AI, LLM tooling, and AI infrastructure. My work blends research depth with production discipline.
Building AI systems that stay fast, reliable, and human-centered.
I work across the stack from model behavior and retrieval pipelines to compiler and systems-level optimization. My interests include LLM quality, latency-aware architecture, and deployment-ready experimentation.
LLM Evaluation and Tooling
Designing practical evaluation signals and constrained decoding strategies for safer, higher-quality outputs.
AI Infrastructure
Production-minded architecture with deterministic testing, observability, and low-latency delivery in mind.
Where research and engineering meet.
Research
Static analysis, secure ML systems, and optimization-driven AI workflows.
Engineering
Safety-critical software architecture, reproducible test frameworks, and scalable integrations.
Writing
AI commentary, movie reviews, book critiques, and mindful art explorations.
Featured Build
Speech-driven Customer Service Assistant
End-to-end STT -> RAG -> TTS assistant designed for low-latency, context-aware dialogue with phoneme-aware constrained decoding for stronger speech quality.
Experience
Selected Industry Experience
Software Engineer - JLR TBSI
- High-precision localization using factor graph optimization.
- Deterministic software architecture for safety-critical behaviors.
- Scalable Software-in-the-Loop testing and validation workflows.
Research Intern - Frontend and Micro-Frontends
- Prototyped Module Federation and monorepo hosting to reduce bundle size and integration overhead via code-splitting.