Udhay 1

I build practical, provider-agnostic LLM tooling that reduces user friction and operational mistakes. In my projects, I focus on clean interfaces, reliable routing across model providers, and engineering choices that keep pipelines consistent.

I created a language-agnostic CLI that supports 15+ LLM providers via LiteLLM, with SHA-256 persistent JSON caching keyed on error fingerprints to avoid redundant API calls. Packaged as a pip/uv-installable cross-platform tool, it lets users switch backends via a single config flag without changing how they use the CLI.

I’ve also engineered multi-agent systems and interactive experiences—like a four-agent CrewAI resume analyzer that performs parsing, JD extraction, gap analysis, and recommendations, then renders structured results through a Gradio UI. I wired inference through LiteLLM so the pipeline stays provider-agnostic across Claude, GPT-4, and open-source models.

During my Generative AI internship at IIT Jammu, I configured multi-step n8n workflows for data and notification pipelines, integrated LLM automation, and deployed a CrewAI resume analyzer that processes candidate profiles with structured gap analysis and role recommendations. I’m motivated by prompt engineering that improves instruction clarity and output consistency across multi-agent pipelines.