Bryan
Pineda.

Self-driving cars have to guess where everything around them is about to go, fast, with partial information. I wanted to actually build a model that does that, not just read papers about it. I trained a deep learning pipeline on nuScenes, real autonomous-driving sensor data, going from raw inputs straight to predicted waypoints; most of the time went into getting the loss function to reflect what "good" actually looks like, not just low error.

Ended up with a model that holds up against the nuScenes benchmarks, and a much better gut sense for why self-driving is such a hard problem.

A bioprinting lab had 3D scans of vascular structures that only made sense as raw volumetric data. Nobody without a CS background could actually look at them. So I built a GUI that lets researchers rotate, slice, and inspect 3D vessel models directly, instead of parsing point clouds by hand.

Researchers started using it day-to-day, which, for a student side project, was the whole point.

I wanted to ask my girlfriend Daniela out, and a text felt like it undersold the effort, so I built a small site from scratch just for her: its own copy, layout, and probably more attention to detail than something only one person would ever see really needed.

She said yes. Still the best-received launch I've shipped.

Most note-taking apps are built for typing, not drawing, and the ones that support sketching rarely feel good to share or collaborate in. I'm building a mobile app for collaborative sketching and note-taking, with a real-time sync layer so two people can draw on the same canvas without it feeling laggy.

Still in progress. Current builds already support live shared canvases; UI polish is next.

I wanted to actually understand how an ECG works at the circuit level, not just know the textbook diagram, so I designed and built the amplification and filtering stages myself. Spent a while figuring out why my "clean" signal kept picking up 60Hz noise from everything else in the room.

Ended up with a circuit that reliably reads a real heartbeat, and a much better intuition for analog signal chains.

Wanted hands-on proof that the structural analysis math I'd been doing on paper actually holds up. I designed a truss structure and iterated on the geometry, pushing for the best strength-to-material ratio I could justify with the load calculations.

A design that held up to the numbers, and a better feel for why engineers obsess over triangles.

Wanted to build something physical and interactive from scratch, not just the display, but the electronics and the game logic driving it too. I designed and built a custom LED board, wiring the hardware and writing the control logic myself.

It plays Pong. Genuinely one of the most fun things I've built.

The classic beginner electronics build: a box whose entire function is turning itself off. I built the mechanism and control logic, then kept going past the obvious version and added a few extra personality modes, including one that acts "shy."

A fully working useless machine that's somehow more entertaining than most of my useful projects.

Wanted to draw my friend's dog, Coco (we call him Cook), properly, instead of the quick sketches I was used to. My first real attempt at a realistic portrait, and I spent most of the time on fur texture, getting the lighting to actually read as fur and not fuzz.

First finished portrait-style piece I was genuinely proud of.

I'd just read a research paper on schizophrenia and had a lot I wanted to process that wasn't going to come out as a sentence, so I drew it instead, leaning into fragmentation and distorted perspective to get at something the paper's language couldn't.

The piece that convinced me drawing and engineering aren't that different; both are just ways of making an idea visible.