Search

Search work, writing, and fragments.

edition / 01/26

selected work / notes / experiments

Software engineer / systems, tooling, and verification

I build software that is clear to use and solid underneath.

I work on software where the surface and the internals both matter: study tools, privacy-heavy systems, verification paths, and machine learning with a clear job to do.

  • currently Computer engineering student / Kathmandu University
  • care about Software that earns trust quickly and still makes sense once you look underneath.
  • on github shuv-amp ↗

selected work

2026

Protocol tooling / correctness and verification work

SP-DIFFER

A correctness harness for BIP 352 Silent Payments implementations, with differential adapters, replayable regressions, and release-verification lanes.

Maintainer-acknowledged upstream bug reports across multiple implementations, with one reduced case later folded into merged BIP-352 test-vector coverage.

  • C++
  • Python
  • Rust adapters
  • Go adapter integration
  • Make
  • Spec-driven docs

2025

Full-stack software engineering

TestLoom

An exam-prep platform that turns scattered study material into a more usable learning system with OCR, adaptive practice, and collaborative tools.

Open-source exam-prep platform connecting OCR ingestion, structured question banks, practice flows, analytics, and discussion in one system.

  • Nuxt 3
  • Vue 3
  • Express 5
  • Node.js
  • MongoDB
  • TypeScript
  • Tailwind CSS

2025-26

Computer architecture / verification

AXIOM-4

A 4-bit RISC-inspired microprocessor with banked memory, a custom ISA, and a verification workflow built in public.

Custom ISA, assembler tooling, and a simulation-first verification workflow that make the architecture testable rather than decorative.

  • Verilog 2005
  • Icarus Verilog
  • GTKWave
  • Python

writing

what stays consistent

The domains can change. The standard does not: understand the problem fully, choose the shape with restraint, and make the engineering stand up in real use.

Note 01

Make the product easy to trust before trying to make it impressive.

Note 02

Let the architecture do the heavy lifting quietly.

Note 03

Keep choosing work that stretches judgment, not just output.