Skip to content
View aashish254's full-sized avatar
💭
Lazy
💭
Lazy

Highlights

  • Pro

Block or report aashish254

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
aashish254/README.md

Aashish Mahato

Systems Architect · AI Integration Engineer · Kathmandu, Nepal

I build local-first tools for people who work with AI agents, and I send the fixes back upstream instead of keeping them in my own fork. Most of my open-source work goes into Laya, an AI decision engine that answers with a single forward pass rather than generating token by token, so speed and correctness are the product; I have worked on its batching, its TypeScript SDK, its CLI, its MCP server and its benchmarks.

My own tools — burn, limen and agentvault — run on your machine with no account and no telemetry; agentvault only talks to a channel you configure yourself, for approvals.

Contributions

Laya — where most of my work lands, and its biggest contributor outside the owner. Batching in the ONNX runtime · batching in the TypeScript SDK · a throughput form of structured decisions · long-document answers in routed windows · the hooks lifecycle ported to TypeScript · the CLI · batch tools over MCP for agents · a schema-driven LangChain decision node · benchmarks pinned to the commit they scored · /health that says where inference really ran · each checkpoint verified against its own SHA-256 map.

freeCodeCamp — curriculum fixes where a hint or an assertion taught the wrong thing: hints that never tested their own cases · a challenge described as an image when it teaches a Markdown link.

public-apis — dead API links removed · the trailing empty cell that broke the Animals table · Auth and HTTPS values normalised to the documented vocabulary.

MDN — typos in the CSS animation shorthand reference · Safari 27 supporting sizes="auto" on image elements.

odysseus — an empty write that would truncate a file · a blank memory_id refused on edit and delete.

headroom — the savings tile priced instead of printing $0.00.

archify — float rounding that lost a node at the solver's exact minimum.

flake8-bugbear — B020 flagging names bound in nested destructuring.

openworker — a crash on a model id that wasn't in its list.

orca — local forge CLIs recovered from broken PATH shims.

ProtocolCanary-Fixtures — recursive fixture discovery in nested directories.

Every link above is a merged pull request. The full set is one search away.

Elsewhere

LinkedIn · LeetCode · All my pull requests

Pinned Loading

  1. ai-image-detection ai-image-detection Public

    🔍 Multi-Modal Forensic Analysis Platform — Detect AI-generated images using frequency-domain math, vision AI & 7 novel detection algorithms

    TypeScript 6

  2. Aniflow Aniflow Public

    🎬 Autonomous AI Studio for Manhwa & Webtoon Recaps — Vision Pipeline (YOLOv8 + Magi v2), Multi-Voice TTS & Video Synthesis

    Python 14 2

  3. burn burn Public

    Local-first, zero-dependency cost ledger for your AI coding agents. Reads the transcripts Claude Code / OpenCode / Codex / Gemini CLI already leave on disk — spend per repo, model, day, and git bra…

    JavaScript 12

  4. limen limen Public

    Local, opt-in proxy for coding agents: measures what your agent re-sends, compiles the scaffolding out of the prompt, and enforces nothing until you say so.

    Python 11

  5. stratmate stratmate Public

    🎮 StratMate — Strategy calculator for mobile gaming alliances with smart resource planning

    JavaScript 11

  6. agentvault agentvault Public

    Runtime permission firewall for AI agents - YAML policies, one-tap approvals (macOS popup/Telegram), tamper-evident signed audit log. Benchmarked: 103/103 attacks blocked, 8/8 normal dev ops allowed.

    Go 10