chris@schmid:~$
← ./home

Blog

Case studies and notes on data engineering, AI, and full-stack builds.

article

I Simulated a Fly's Escape Reflex on Its Real Wiring, Then Started Cutting Neurons

Google/HHMI Janelia just mapped a fruit fly's entire nervous system down to the synapse. I turned its escape circuit into a spiking simulator and measured exactly how many neurons you can silence before the fly stops dodging — almost none.

PythonNeuroscienceConnectomicsSimulation

case study · personal project

Jersey City Bikeshare

Production-style data platform over 230K Citibike trips (Jersey City) — dbt medallion architecture, Kafka streaming, Terraform-provisioned Snowflake, and Airflow orchestration.

SnowflakedbtAirflowKafkaTerraform

case study · personal project

NYC Taxi Data Pipeline

320M NYC yellow taxi trips across 7 years, processed through a Bronze/Silver/Gold pipeline on Snowflake and dbt with a quarantine layer gating data quality — total compute cost: $9.94.

PythonSnowflakedbt

article

An On-Chain Factory for One-Transaction NFT Collections

A Solidity factory contract that deploys a fresh ERC-721 collection, mints the first token, and hands over ownership — all in a single transaction, verified on Sepolia and wired up with wagmi.

SolidityOpenZeppelinERC-721wagmi

article

I Tried to Predict Apple's Stock Price With Linear Regression. Here's Why You Shouldn't.

A backtest showed 0.78 R² and 77.5% directional accuracy. Then I ran it on real future data and it barely beat a coin flip.

PythonMachine LearningFinance

article

World ID SSO: On-Chain vs. Cloud, Side by Side

A Bachelor's thesis project comparing two ways to let a user prove they're a unique human — a Solidity contract verifying a World ID zero-knowledge proof on-chain, and a NextAuth OIDC flow doing the same in the cloud.

Next.jsBlockchainWorld IDOAuth

article

A Vanilla GAN on MNIST, and What FID Actually Told Me

A fully-connected generator and discriminator trained for 500 epochs on MNIST — plus the numerical instability FID hits at small batch sizes, and how I dealt with it.

PyTorchPythonMachine LearningComputer Vision