Devanshu User

I’m a Machine Learning Intern focused on turning LLM ideas into production systems that teams can rely on. I’m currently pursuing a B.Tech in Computer Science & Engineering at IIIT Una, and my work blends ML engineering, backend development, and evaluation rigor.

At Planto.ai, I shipped a production VSCode coding assistant with features like code completion, inline documentation, and multi-file context—deployed to 3 enterprise clients and cutting context-switching by ~30%. I owned CI/CD pipelines using GitHub Actions to reduce deployment time by ~40% and ensure reproducible builds, while also fine-tuning a proprietary LLM on 500+ curated instruction-response pairs for a ~15% accuracy gain on domain benchmarks.

I also architected end-to-end RAG workflows (chunking, embeddings, Pinecone indexing) that reduced retrieval hallucinations by ~25%, gathered requirements from 5+ enterprise partners, and used A/B testing to shape internal best practices. Beyond internships, I build full-stack AI agents—like a dual-channel voice + web RAG portfolio agent—and explore applied forecasting with NLP-enhanced volatility models, always with measurable outcomes and fast iteration.