Ayush Ranjane Portfolio
Three real deployed ML systems — a CO2 compliance platform, an AI task-scheduling copilot, and an AQI forecaster — each with a stated architecture decision (Prophet over ARIMA, XGBoost over Random Forest) and a live demo link, not just a GitHub repo.
Location
Pune, India
Skills
Notable
- 20 completed Coursera/IBM certificates (Jul 2023 – Mar 2024) including the full IBM Data Science and IBM Machine Learning professional certificate tracks, each linked to a viewable PDF
- Three deployed, publicly accessible projects with separate live-demo and source-code links (GreenCO2, LifeSaver, AQI Prediction System)
- Active GitHub (github.com/Ayush-Ranjane) with repos matching the portfolio's named projects (GreenCO2, AQI-Prediction-Analysis-System, Titanic-Survival-Prediction)
What makes it work
A breakdown of the choices that make this portfolio stand out.
"Architecture Decisions" callout on every project card
Instead of a features list, each of the three case studies has a labeled box explaining a specific tradeoff: "Chose Prophet over ARIMA for its superior handling of missing emission logs and weekly seasonality" and "Deployed XGBoost over Random Forest for gradient-based optimization on non-linear pollutant correlations." This is the single strongest element on the page — it shows reasoning, not just a tool list, which is exactly what separates a tutorial-follower from someone who can defend a choice in an interview.
Metrics instead of adjectives
Each project leads with three hard numbers in a stat row — GreenCO2 shows 94% forecast accuracy, <50ms alert latency, 100% compliance; the AQI system shows a 0.92 R² score and <2.5 RMSE across 11 pollutants. A junior candidate quoting an R² score and RMSE instead of "built a machine learning model" reads as someone who actually validated the model rather than just fit one.
Full-stack framing of an ML portfolio
All three projects pair a model (Prophet, Isolation Forest, XGBoost) with a served product — FastAPI/Flask backends, a React or Next.js frontend, and a live Vercel/Render deployment link plus a separate GitHub source link. Most student ML portfolios stop at a Jupyter notebook; this one ships the notebook's output as something a recruiter can click into and actually use.
20 stacked Coursera/IBM certificates with completion dates, not a badge wall
The certificates section runs July 2023 through March 2024 in a coherent sequence (Python for Data Science → EDA → Supervised/Unsupervised ML → ML Capstone → Generative AI), each with a one-line description of what was actually done and a link to the PDF. It reads as a genuine self-paced curriculum completed in order, not a scattershot of credentials collected for the sake of a long list.
What ML Engineers can take from this
Specific, actionable tips to apply to your own portfolio — no generic advice.
The "Deep Dive" links on each project card go to dedicated /projects/[slug] pages — that structure is doing real work; make sure each deep-dive page front-loads the same problem/decision/metric framing used on the card, since that's what a hiring manager will actually click into.
The bio still opens with "Aspiring AI & Machine Learning Engineer" right above three production-style case studies with live traffic-serving demos — drop "aspiring" once the projects themselves are this concrete; the hedge undersells the metrics that follow it.
GreenCO2's "100% COMPLIANCE" and LifeSaver's "100% SYNC ACCURACY" read as suspiciously round for a real metric — a specific number (or a stated sample size behind the percentage) would be more credible than a clean triple zero.
The GitHub account (github.com/Ayush-Ranjane) only shows 3 pinned repos matching the site's projects with 1 star each — fine for a student portfolio, but worth adding a README to each repo with the same architecture-decision framing so the GitHub side stands on its own, not just as a link target from the portfolio.
More ML Engineer portfolios
Ready to build your portfolio?
Follow the ML Engineer roadmap — skills, projects, and timeline to get hired.