Open to internships & new-grad roles, 2027

Muhammad Hasham Hussain

Data Science & Machine Learning

M.S. student building projects across machine learning, statistical analysis, and data visualization.

View ProjectsContact

About

Hi, I'm Hasham, welcome to my portfolio. I'm currently pursuing an M.S. in Data Science & Analytics, working on projects across machine learning, statistical analysis, and data visualization. I'm looking for internships (fall, spring, and summer) as well as full-time / new-grad roles starting in 2027, feel free to reach out.

Projects

Keeper — Songwriting Voice Memo Organizer

Keeper showing a voice memo waveform with takes labeled Verse 1, Unlabeled 1 and Bridge 1, spoken commands marked beneath, and a takes list beside the song canvas

Finds the takes in long songwriting voice memos and files them by what you said. An Audio Spectrogram Transformer on a Modal GPU separates singing and playing from speech and silence, then Meta's Muse Spark reads the spoken commands around each take ("file that as a verse", "scrap that") and sorts takes into sections, keeping scrapped ones in a Rejected folder. Takes can be arranged on a song canvas, played through like a rough demo, and exported as a WAV or stems. Built solo for the Meta Global AI Developer Hackathon; hand-checked on one demo memo (5 of 6 takes found, 4 of 5 commands handled).

  • Python
  • FastAPI
  • Modal
  • Audio Spectrogram Transformer
  • Muse Spark
  • Web Audio API
  • JavaScript

shadenav — Shade-Aware Pedestrian Routing

shadenav map showing a shortest route and a shadier alternate route between two campus points, with a sun-exposure comparison panel

A routing tool that models where building shadows fall at any date and time, scores every street segment by sun exposure, and finds walking routes that trade a little extra distance for a lot less time in direct sun. Built on an end-to-end geospatial pipeline over the Chico State campus: OpenStreetMap street networks and building footprints, sun-position modeling, shadow-casting geometry, and shade-weighted shortest-path routing. Validated across 200 routes — a mild shade preference cuts sun exposure by roughly 20% for about 2% extra distance.

  • Python
  • pvlib
  • NetworkX
  • OpenStreetMap
  • Geospatial

Keyword vs. Meaning Search

A full-stack search engine comparing BM25 (keyword-based ranking) against neural embeddings (semantic ranking) over 500+ real Spotify tracks. Includes live Spotify search integrated as a retrieve-then-rerank layer, with automatic fallback to a precomputed local corpus if the live call fails or times out.

  • Python
  • FastAPI
  • onnxruntime
  • rank_bm25
  • Spotify Web API
  • JavaScript

Does Song Energy Predict Popularity?

Analyzed 25,736 Spotify tracks across six genres to test whether a song's energy predicts its popularity. Used Pearson correlation and linear regression with an energy × genre interaction term. Found a statistically significant but very weak negative relationship — energy explains under 1% of the variance in popularity, and it does not meaningfully change across genres.

  • R
  • tidyverse
  • ggplot2
  • Linear Regression
  • Tableau

Skills

Languages

  • Python
  • SQL
  • R

ML / Data Science

  • scikit-learn
  • PyTorch
  • pandas

Visualization

  • Matplotlib
  • Tableau
  • Plotly

Contact

Open to data science and ML roles. Feel free to reach out.