Skip to content
View noahchalifour's full-sized avatar

Block or report noahchalifour

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. 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
noahchalifour/README.md

Noah Chalifour

Senior Software Engineer at Regie AI, previously at AWS (RDS). I build AI agents and the infrastructure they run on. Based in Vancouver, BC.

  • At Regie AI: core engineer on Regie Go, an AI sales agent built from scratch: onboarding state machine, BullMQ and SSE pipelines, citation-validated email drafting, voice onboarding.
  • At AWS: distributed systems for Amazon RDS, including the Global Writer Endpoint and a cleanup that freed 10 PB of storage, plus internal LLM tools on Bedrock.
  • On the side: agent tooling, a Kubernetes home lab, and products at chalifour.dev.

Agent infrastructure

Project What it does
berth A lease-based pool of self-hosted Android emulators. Coding agents check out a phone, tablet or TV over MCP, humans watch and take over live. Python, Kubernetes, Prometheus.
dsh-anthropic-request-size-guard Keeps long, screenshot-heavy agent sessions under the Claude API's 32 MB request limit (a stuck 38 MB request became 6.1 MB) without breaking the prompt cache.
dsh-anthropic-image-clamp Downscales images so many-image Claude requests never hit the 2000px limit. Deterministic, cached by content hash.
homeassistant-for-openclaw Gives an AI agent live smart-home context and device control through Home Assistant, with a three-tier read, watch and write access model.

Machine learning

Project What it does
rnnt-speech-recognition End-to-end streaming speech recognition with RNN Transducers in TensorFlow 2, after Google's on-device RNN-T paper. 250+ stars.
memn2n End-to-end goal-oriented dialog with memory networks.

Published research: Question Answering Artificial Intelligence Chatbot on Military Dress Policy (Department of National Defence), fine-tuning BERT-based models for domain-specific QA.

Stack

Python, TypeScript, Java, Go · React, Node.js, Flutter · AWS (Lambda, Bedrock, RDS), Kubernetes, Terraform, Argo CD · LLM APIs, MCP, RAG

Links

noahchalifour.com · chalifour.dev · LinkedIn · Resume

Pinned Loading

  1. rnnt-speech-recognition rnnt-speech-recognition Public archive

    End-to-end speech recognition using RNN Transducers in Tensorflow 2.0

    Python 250 79

  2. berth berth Public

    Lease-based Android emulator pool: MCP for agents, live view for humans

    Python

  3. dsh-anthropic-image-clamp dsh-anthropic-image-clamp Public

    DeepSeek Harness plugin: downscale images on Claude requests so many-image requests never hit the 2000px limit

    JavaScript

  4. dsh-anthropic-request-size-guard dsh-anthropic-request-size-guard Public

    DeepSeek Harness plugin: shrink oversized Claude requests so they never fail with HTTP 413 request_too_large

    JavaScript