SanidhyaKumar.
I build software at the intersection of artificial intelligence, full-stack engineering, and developer experience.
Beyond the résumé.
I'm a fourth-year Computer Science and Engineering student at PES University building intelligent software systems. Over the last few years, I've explored everything from full-stack web development to machine learning, distributed systems, and AI infrastructure. Rather than treating AI as a simple add-on, my projects focus on designing systems like adaptive RAG pipelines and multi-agent platforms where machine learning meaningfully improves the core product.
I enjoy solving engineering problems where architecture matters just as much as algorithms. Whether it's reducing database latency, optimizing semantic search, or building tools that make developers more productive, I'm most motivated by creating software that's reliable, efficient, and genuinely enjoyable to use.
The tools I build with
A working set of languages, frameworks and platforms, filter by category to see how they group, or browse all 26 at once.
Most Used Tech Stack
Some of My Work
ThingsI've made.
- 01
Doxa
AI-powered knowledge and reasoning platform. Instead of simply querying an LLM, Doxa combines retrieval, reasoning, verification, and intelligent orchestration to produce answers grounded in reliable information.
PythonReactPostgreSQL2026 - 02
Adaptive Recursive RAG
Research-driven retrieval pipeline for complex question answering. A retrieval architecture that dynamically decides whether more retrieval is actually necessary instead of using a fixed number of retrieval steps.
PyTorchHuggingFaceFAISS2026 - 03
LC Helper
LC Helper is a privacy-focused, recall-first revision system designed to ensure you never forget an algorithm you have solved. It combines automated capture via a browser extension, strict AI explanations using Google Gemini, and a smart spaced repetition algorithm into a premium, distraction-free dashboard.
Next.jsReactTypeScript2026 - 04
Argus-PG
PostgreSQL query optimization and index validation engine. Argus validates database optimizations before they ever reach production by generating indexing hypotheses and benchmarking improvements.
PythonPostgreSQLDocker2026 - 05
SnapBack
Version snapshots without Git. SnapBack is a command-line tool that lets developers create lightweight project snapshots and restore previous versions, using content-addressable storage for speed.
TypeScriptNode.jsCLI2025 - 06
LocalSearch
LocalSearch is a native macOS desktop search application with a Rust search/index core, a SwiftUI/AppKit frontend, and a C-ABI FFI bridge between the two.
RustSwiftUImacOS2026