Building Your Portfolio

📘 Learn Building Your Portfolio from zero

Start from zero: a portfolio is a curated, indexed collection of evidence that you can do the job. Think of it like the menu and photos outside a restaurant. A hungry person (the hiring manager) walks past dozens of doors. They will not cook a sample meal at every one. They glance at the menu (your sections) and the photos (your projects) and decide in seconds whether to step inside (interview you). Your job is to make the menu legible and the photos appetizing and honest.

This breaks into three first-principles tasks. (1) Sections are the menu's structure: About (who you are), Projects (proof), Skills (capabilities), Contact (how to hire you). Each answers one question the reader is already asking. (2) Project descriptions are the photos with captions. A good caption leads with the result, then fills in a fixed schema - Problem, Role, Approach, Result. This is an adaptation of the STAR framework from behavioral interviews (Situation, Task, Action, Result); the same evidence that structures a strong interview answer structures a strong project blurb. (3) AI enhancement is a kitchen assistant that helps you plate a dish you already cooked - it does not invent ingredients.

Worked example. Raw input: "Built a website for a club." Apply the schema. Problem: the club tracked 200 members in a spreadsheet that broke during signups. Role: sole developer. Approach: built a React + Firebase app with a self-service registration flow. Result: cut signup time from ~10 min to ~30 sec and eliminated double-bookings. Output: "Designed and shipped a member-management web app (React, Firebase) that replaced a failing spreadsheet for 200 members, cutting signup time ~95% and eliminating double-bookings." Same project, far more signal.

Key insight: a portfolio is an optimization problem - maximize signal per second of reader attention. Curate ruthlessly, lead with the result, and quantify every claim you can defend.

✨ Added by the guide to build intuition — not from the source course.

🎯 Guided practice

1. Easy: Pick the core sections for a new-grad SWE portfolio.

   Step 1: List the questions a reader has, in order: "Who are you? Can you prove you can build? What tools do you know? How do I reach you?"

   Step 2: Map one section to each question, one responsibility each (high cohesion): About, Projects, Skills, Contact.

   Step 3: Order by reader priority. Projects are your strongest proof, so place them prominently near the top after a one-line About. Drop anything that does not answer a reader question (a hobbies wall, a visitor counter) - that is noise.

   Result: a four-section structure ordered About -> Projects -> Skills -> Contact. The pattern: each section earns its place by answering exactly one reader question.

2. Medium: Turn a weak project bullet into a strong description, then enhance with AI safely.

   Input bullet: "Made a machine learning model for predicting house prices."

   Step 1: Diagnose - it states activity, not outcome, and has no metric. It is "I worked on X."

   Step 2: Extract the Problem/Role/Approach/Result fields by interrogating yourself. Problem: realtors priced listings manually, with high variance. Role: I built the model and ran the evaluation. Approach: trained a gradient-boosted regressor on 10k listings with feature engineering on location and size. Result: reached a mean absolute error (MAE) of about 8%, versus a manual baseline of around 20%.

   Step 3: Compose, leading with the result: "Built a gradient-boosted price-prediction model on 10k listings, reaching ~8% MAE versus a ~20% manual baseline (Python, scikit-learn)."

   Step 4: Enhance with a specific, grounded prompt, never an open one. Good prompt: "Here is my project bullet and the target job description. Rewrite for clarity and impact, keep every number exactly as given, do not invent metrics or technologies, and match the JD's tone." Then verify: confirm the model changed no numbers and added no claim you cannot defend in an interview.

   The pattern: structure first (lead with the metric, fill the schema), AI second as a constrained editor you fact-check - never the other way around.

✨ Added by the guide — work these before the full problem set.