Skip to content
View MLGalusha's full-sized avatar

Highlights

  • Pro

Block or report MLGalusha

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.

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
MLGalusha/README.md

Mason Galusha

Applied AI Engineer building with LLMs, retrieval, structured extraction, and coding agents.

Portfolio · LinkedIn · Email

Based in Raleigh, NC. Open to Applied AI Engineer, AI Backend Engineer, RAG Engineer, agent tooling, and full-stack AI roles.

Featured work

Construction PDF search and cited-answer tool for long technical documents where exact evidence matters.

  • Parses PDFs, chunks by page and section context, retrieves ranked evidence, expands nearby source context, and answers with page-level citations.
  • Supports lexical, BM25, vector, hybrid, and routed retrieval modes with query planning for follow-ups, multi-part questions, page-anchor questions, and clarification cases.
  • Includes evaluation and debugging tools for retrieval quality, citation verification, answer quality, query planning, conversation behavior, feedback conversion, latency, and UI run logs.

Public historical letter archive built to make scanned family letters searchable, readable, and publishable as the collection grows.

  • Public site for browsing, search, collections, people, places, and letter pages.
  • Admin system for upload, review, transcription, structured metadata extraction, entity resolution, usage tracking, notifications, and content tools.
  • AI handles the slow parts, but human review gates keep control before content goes public.

Second Braincell

Private local Node.js CLI that lets coding agents use ChatGPT for document review, multi-turn reasoning, file/PDF retrieval, image generation, and Deep Research jobs.

  • Replaces browser-automation-heavy workflows with direct request/response paths.
  • Stores local job artifacts, supports model fallback logic, manages Project instructions, and checks auth/session state.
  • Built for workflows where implementation happens in a repo but useful context lives in ChatGPT.

Public repos

  • SonicGen: Python audio fingerprinting and dedup engine using spectral peak detection, constellation hashing, and alignment-offset matching.
  • PianoTranscriber: PyTorch model that transcribes piano audio into MIDI from spectrogram inputs.
  • Staffclaw: full-stack scheduling and operations dashboard built from a reverse-engineered workplace scheduling API.
  • JobTracker: agent-native job search pipeline for role intake, company research, resume tailoring, cover letters, and interview prep.

Focus

  • Retrieval-augmented generation and cited answers
  • Long-document search and document intelligence
  • Structured extraction and human review workflows
  • Agent tooling and coding-agent workflows
  • Python, TypeScript, React, Node.js, PostgreSQL, OpenAI API, GCP

Pinned Loading

  1. letter-archive letter-archive Public

    Live digital archive for historical letters. AI vision transcription with human review.

    TypeScript 1

  2. job-tracker job-tracker Public

    Agent-native personal job-search tracker. Event-sourced, driven by Claude Code.

    TypeScript 1

  3. PianoTranscriber PianoTranscriber Public

    An AI-powered tool for transcribing solo piano performances from MP3 files into MIDI and sheet music using deep learning.

    Jupyter Notebook 8

  4. SonicGen SonicGen Public

    Hand-written audio fingerprinting engine for duplicate detection and source matching.

    Python 1

  5. portfolio-website portfolio-website Public

    Personal portfolio site. React 19, TypeScript, Vite.

    TypeScript 1