BD

AI software engineer · Product builder

Binaya
Dhakal.

I build AI products and developer tools.

AI Software Engineer at Fluid, building production commerce infrastructure, payments, attribution and AI features at scale. I also create developer tools and mobile products, and publish machine-learning research.

View selected work Get in touch
Portrait of Binaya Dhakal

01 — Selected Work

Current role · Since 2021

AI Software Engineer at Fluid

Active

Building the core platform for modern direct-selling commerce.

One of Fluid's earliest engineers, I have worked on the product since 2021—first through Danphe and directly with Fluid since 2025. I build across the production stack, from storefront infrastructure and payment systems to affiliate attribution and AI-powered rep outreach.

  • Commerce foundationLiquid theme engine and multi-tenant storefronts
  • Payments & subscriptionsStripe, Braintree and VGS tokenization
  • Growth systemsFairShare attribution and AI rep outreach
  • Production scaleBrands exceed $1M on peak sales days
fluid.app

Featured project · Co-creator

NVIDIA Inception

Yanib

The shipping layer for software teams.

An independent project I co-created to turn commits into AI-drafted release notes, review them against the actual code, and publish them to GitHub, Slack, Discord and email. Public shipping profiles make developer work visible. Live since March 2026.

yanib.dev

  • Live productShipping since March 2026
  • NVIDIA InceptionProgram member
  • Built in publicYanib tracks its own releases
View my live shipping record

Built with Yanib and dogfooded on Yanib itself. If the live record does not load, view it on Yanib

Creator

Capsify

Save it. Share it. Open it together.

Collaborative capsules of photos, places, notes and songs — built with friends, sealed until reveal day.

iOS · Android   capsify.app

Creator

Solmari

A second brain that never leaves your phone.

Private, offline, on-device AI — chat grounded in your own notes. No account, no servers, no tracking.

iOS   solmari.app

Creator

SpeedCube AI Analyzer

Train smarter, not just faster.

An AI solve coach for speedcubers — record a solve and get phase splits, a pause map and coaching feedback, on a WCA-compliant timer.

iOS   speedcubeaianalyzer.com

02 — Journey

  1. May 2025 — Present

    AI Software Engineer, Fluid

    One of Fluid's earliest engineers since 2021. Built major production surfaces across the Liquid theme engine, payments and subscriptions, FairShare attribution, and AI-powered rep outreach; joined Fluid directly in 2025.

  2. 2026 — Present

    Built and launched Yanib

    Independent project · live since March 2026 · NVIDIA Inception Program · alongside shipping Capsify and Solmari.

  3. Aug 2024 — Jan 2026 · Graduated

    M.S. Data Science, University of New Haven

    Research Assistant · Provost's Scholarship (75%, paid stipend) · deep learning on microbiome data — autoencoders & VAEs for mortality prediction, published as first author.

  4. May 2021 — May 2024

    Senior Software Engineer, Danphe Software Labs

    Full-stack & applied ML — cyber-attack detection with SVM, CNN, RNN & NLP. Fluid was a client project here; I was one of the first engineers on it.

  5. Dec 2016 — Jan 2021

    B.E. Computer Engineering, Paschimanchal Campus (IOE, Tribhuvan University)

    Tribhuvan University Merit Scholarship ×3 · winner, Capture the Flag security competition · national #2 Rubik's cube speedsolver (WCA).

03 — Research

First author · Peer-reviewed

Integrating multi-compartment microbiome data with clinical parameters enhances mortality prediction using autoencoder

Dhakal B, Savarapu LSK, Sayed K

Journal of Microbiological Methods · Vol. 238, 107267 · November 2025

Deep representation learning across multiple body-site microbiome compartments, fused with clinical parameters, to predict patient mortality — the core of my research as a Research Assistant at the University of New Haven.

DOI PubMed

Effects of Sudarshan Kriya Yoga (SKY) on the stress and self-esteem of medical doctors in a tertiary care hospital

Co-author · Annals of Medicine and Surgery · sentiment analysis with neuro-medical researchers.

2024 · DOI

Google Scholar  ·  Writing — the engineering blog

04 — Open Source Tools

AgentBarrier

Transaction-safe policy, human approval, and execution gateway for consequential AI-agent actions — exact argument binding, atomic execution claims, durable replay, and integrity-linked audit receipts. Why I built it

GitHub · PyPI

tool-call-guard

Deny-by-default security gate for AI agent tool calls — one JSON policy, enforced in JavaScript and Python with a shared audit schema. Launch story

npm + PyPI · GitHub

react-native-local-llm

One interface for every on-device LLM runtime in React Native — Apple Foundation Models, ExecuTorch and Cactus. Extracted from Solmari's production inference layer.

npm

vitest-mcp

Vitest harness and matchers for MCP servers — in-memory transport, real protocol path, millisecond tests.

npm

context-budget

Fit chat history into any model's context window — sliding-window packing with running summaries. Zero dependencies, shipped for JavaScript and Python.

npm + PyPI

ai-sdk-cache

Record/replay cache middleware for the Vercel AI SDK — free dev iterations, offline deterministic tests.

npm

Selected packages across npm and PyPI — open source, typed, and tested in CI. github.com/binaydhakal

Let's build something.

For engineering opportunities, product collaborations, open-source work, and conversations about applied AI.

binaydhakal35@gmail.com