From Non-Profit Ops Manager to Building Neural Networks: Week 1

Source: Dev.to

Introduction

Six months ago I was managing operations for a basketball association—scheduling, budgets, membership data, spreadsheets. It was meaningful work, but I kept watching the AI revolution from the sidelines. I decided to move to the other side.

I’m not starting from zero. Earlier this year I completed a HyperionDev Data Science Bootcamp, graduating first in my class. I can wrangle data, build basic ML models, and deploy them, but data analysis and deep AI development are very different beasts. My goal is to work at the frontier of AI development within the next 5‑6 years—building training environments, agent systems, and shaping the direction of the technology.

Week 1 Projects

House Price Predictor

• Full end‑to‑end ML web app.
• Real dataset (545 housing records).
• Data‑cleaning pipeline, model selection (Linear Regression vs. Random Forest; Linear Regression performed best).
• Deployed live on Streamlit Cloud – you can try it now.

Sales Analytics Dashboard

• Streamlit‑based sales analytics app with production‑grade architecture.
• Real dataset (Superstore, 545 records).
• Robust data‑validation pipeline (multi‑delimiter/encoding support).
• 8 interactive visualizations, dynamic filtering.
• 94.9 % test coverage (unit, integration, UI tests).
• Deployed live on Streamlit Cloud – upload your own CSV or explore the sample dataset.

Neural Network from Scratch

• Implemented a perceptron with sigmoid activation and gradient descent using only NumPy (no PyTorch or scikit‑learn).
• Tested on AND, OR, and NAND logic gates, then applied to a digit‑recognition task.
• Watching the loss curves drop as the weights updated was one of the most satisfying moments of my learning journey.

Fast.ai Practical Deep Learning – Lesson 1

• Completed the first lesson of Fast.ai’s free Practical Deep Learning course.
• Trained my first image classifier.

Reflections on the AI Landscape

Deep learning feels like a rabbit hole with no visible bottom. Beyond it lie reinforcement learning, multi‑agent systems, distributed training infrastructure, interpretability research, and alignment work—each a career’s worth of depth. The field is larger than any single discipline I’ve encountered and is expanding faster than anyone can fully track. Rather than being intimidating, this breadth feels electric.

Six months ago I was building financial models in Excel; this week I implemented backpropagation by hand. The pace of what’s possible when you commit fully to learning is genuinely surprising.

I can now build and deploy ML models, but I still lack depth in reinforcement learning, distributed systems, and research methodology. I haven’t published anything, and I don’t have a computer‑science degree. The people at the frontier are extraordinary, and I’m not there yet—but I’m aiming to be in six years. Week 1 has placed me exactly where I need to be: building foundations, staying consistent, and moving forward every day.

Looking Ahead

Week 2 will focus on:

• Finishing Fast.ai Part 1 and diving deeper into CNNs.
• Implementing a Transformer architecture from scratch.
• Beginning reinforcement learning fundamentals—the area I’m most excited about and intend to specialize in.

I’ll document the journey here, sharing wins, confusions, long‑lasting errors, and those moments when everything finally clicks.

Follow My Journey

• GitHub:
• The House Price Predictor and Sales Dashboard are live on Streamlit Cloud—feel free to explore them.

Week 1 complete. See you next week.