Spike Foundry
File 01
Status: Testing
File 01 · The Lab

Useful projects, not tutorial theater.

Technical tools learned one at a time, then brought together into work that can move a career.

→ Start with Python. Start with the boring thing.

The Lab

The Lab is where Spike Foundry turns useful technical tools into projects that can actually matter at work.

Not tutorial theater. Not another course you half-finish and never mention again. The point is simpler than that:

Build the thing. Understand why it works. Package the proof. Use it to get better at your job, ask for more responsibility, earn a raise, or move toward a better position.

I did not learn technical tools because I loved syntax. I learned them because there was a folder full of junk, a report nobody wanted to clean, a process everyone complained about, or a task I knew a computer should be doing for me.

That is the operating principle here.

The path

We are going one tool at a time.

Each stop has three jobs:

  • learn the tool well enough to use it;
  • build a real artifact with it;
  • explain the business value in plain English.

No fake startup clone. No “hello world” pretending to be a portfolio. Useful projects, built in public, with the messy parts left in.

The first track

1. Python

Use Python to automate the boring part: files, folders, reports, cleanup, repetitive work, and small scripts that save time.

First project: a safe file triage script that scans a messy folder, shows a dry-run plan, organizes files, and writes a summary report.

2. FastAPI

Turn a useful script into a small service with routes, request validation, API docs, and a clean way for other people or tools to use it.

Project: a file/report analysis API.

Store useful information instead of losing it in folders, notes, and one-off chats.

Projects: PostgreSQL, pgvector, and a comparison with Weaviate and Pinecone.

4. LLM fundamentals

Learn what large language models are actually doing at a practical level: prompts, tokens, context windows, embeddings, retrieval, and why confident answers still need verification.

Project: a small document assistant that can answer from source material instead of guessing from vibes.

5. LangChain, LangSmith, Langfuse, and LangGraph

Use the LLM ecosystem without pretending the tools are magic.

Projects:

  • a retrieval chain;
  • traced and debugged model calls;
  • observability with feedback;
  • a state-based workflow that knows when to stop and ask for a human.

6. Machine learning fundamentals

Learn enough scikit-learn and PyTorch to understand the difference between classic machine learning, neural networks, and LLMs.

Project: classify messy work items and compare a simple ML model against an LLM-based approach.

7. Terraform

Deploy the work like it matters.

Project: infrastructure as code for the API, database, and supporting services.

The capstone

The Workbench is not live yet. It is the planned capstone for the Lab once the earlier projects have real artifacts behind them.

The likely pieces are candidates, not a promise: Python ingestion scripts, an API, a database, document search, workflow orchestration, observability, ML fundamentals, and deployment notes. The stack should earn its way in as the projects prove what is actually useful.

The goal is not to impress strangers with a complicated stack.

The goal is to build something useful enough that you can explain it to a manager, a recruiter, or yourself and say:

I found a real problem, built a working system, understood the tradeoffs, and can explain the value.

That is the lab.

Start with Python. Start with the boring thing. Start where the work is already annoying.