jjanella/site
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

css and posts

Browse Source
This commit is contained in:
Jeremy
2026-01-25 20:28:13 -05:00
parent 552cad6b4a
commit a2ce84707d
18 changed files with 264 additions and 173 deletions
Download Patch File Download Diff File Expand all files Collapse all files

64
static/posts/gamedev/network.md
View File

@@ -1,64 +0,0 @@
<h1>Servers & Networking</h1>
<p>
Open source and networking hobbies led me to depoly a set of
interconnected Linux servers:
</p>
<ul>
<li>
<h3>NGINX Reverse Proxy and VPN Server</h3>
<p>
This server is the internets access to all of my other servers,
and the glue holding them together. I was using IPTables briefly
but have moved to NGINX to forward traffic from specified ports
onto private servers connected to it. This device doubles as a <a
href="https://www.wireguard.com/">Wireguard VPN</a
> server which in the past connected servers on the same virtual
network, however I have switched to reverse ssh tunnels to connect
other servers to this one. The VPN is still useful for odd networking
tasks and troubleshooting.
</p>
</li>
<li>
<h3>Lightweight Server</h3>
<p>
Running on Raspberry Pi, this servers main job is to run
Subterstrike servers, Minecraft servers, other game servers, and
this website you're looking at.
</p>
</li>
<li>
<h3>AI & FTP Server</h3>
<p>
A powerful computer with an RX 6950 XT GPU runs my local
generative AI servers including <a
href="https://invoke-ai.github.io/InvokeAI/">Invoke AI</a
>
for Image Generation,
<a href="https://ollama.com/?ref=useaiforit">Ollama</a>
and <a href="https://docs.openwebui.com/">Open-WebUI</a> for LLM
services. As this is also my primary device with a few spare
terabytes of NVMe storage it contains a
<a href="https://en.wikipedia.org/wiki/Vsftpd"
>VSFTPD file server</a
>. By the way, this server runs on
<a href="https://archlinux.org/"><i>Arch Linux</i></a> :)
</p>
</li>
<li>
<h3>Derelict Laptop Promoted to Server</h3>
<p>
An ancient gaming laptop with significantly more computational
performance than physical durability was promoted to server,
where it runs my <a href="https://nextcloud.com/about/"
>Nextcloud</a
>
self hosted office suite. While originally this was a 'because I
can' project, I felt justified when I later saw
<a
href="https://www.digitaljournal.com/tech-science/microsoft-says-u-s-law-takes-precedence-over-canadian-data-sovereignty/article"
>Microsoft announcing USA demands come before Canadian
privacy</a
>.
</p>
</li>
</ul>

53
static/posts/gamedev/phobos.md
View File

@@ -1,53 +0,0 @@
<h1>Phobos</h1>
<p>
One of the coolest things I've developed is a game, fully written in
assembly.<br /><br />
I say this because it is the perfect intersection of many things I enjoy;
low level programming, physics, space, simulation development, and game design.
Also small enough of a project that I could complete it in under a couple
of months, so another win. I even named the game based on the emulator, as
Phobos is a moon of Mars <br /><br />The instructions for this
assignment were very broad, to develop a platformer game in assembly.
This left a lot of room for creative freedom, especially due to the
loose definition of platformer. With infinite time I would have
developed a whole rocket simulator, and used the justification that the
launchpad was a platform.
<br />
<br /> On the technical side this was a very difficult project. There are
only about 16 variables actually usable in MIPS, and a lot of them are reserved
to specific logical controls under best practice. Then following the many
loops and recursive, real-time functionality required extensive and efficient
use of the stack. Ideally accessing a few bytes shouldn't strike the fear
of low framerates to mind, but the Mars-MIPS emulator I was using was far
from efficient.
</p>
<h3>Notable Features</h3>
<ul>
<li>Raycast collision detection and repositioning</li>
<li>Sub-pixel velocity and positioning</li>
<li>Awesome pixel graphics (the rocket shoots fire!)</li>
<li>High framerates (for a low-performance emulator)</li>
<li>Enemies shooting at you!</li>
<li>Loot pickups</li>
</ul>
<p>
Probably the most annoying part of this was that the emulator couldn't
handle simutaineous key inputs. Fortunately I have over a thousand hours
experience landing rockets on the <a
href="https://kerbalspaceprogram.fandom.com/wiki/Mun">Mün</a
>
so I was comfortable giving a few unique key inputs per second. I never imagined
this is something I would be proud, and I still don't think I am. However,
it works.<br /><br />At this point you may have realized I am a bit of a
space nerd. Enjoy this recording I made at 1am demonstrating the
features of my game to a TA, compressed to the point that the audio has
texture.
</p>
<video width="90%" controls>
<track kind="captions" />
<source src="src/lib/phobos/demo.mp4" type="video/mp4" />
Your browser does not support the video.
</video><br />
The
<a href="https://github.com/jjanella/Phobos">source code</a> is also available
for your enjoyment.

2
static/posts/gamedev/post.typ
View File

@@ -1,6 +1,8 @@
#let post_title = "Game Dev"
#let post_preview_image = "flying.webp"
#let post_summary = "Creating an open world multiplayer terrain destruction game in Godot and Rust"
= Game Development
I am currently developing an underground-submarine openworld multiplayer
physics inspired game. As a gamer and enjoyer of large, rich maps and

59
static/posts/gamedev/uav.md
View File

@@ -1,59 +0,0 @@
<h1>Building a UAV</h1>
<p>
I chose to build a drone from scratch for my AP Physics final project.
We were given an open ended assignment to <i>build something cool</i>
and I wanted a drone. My choice was much more ambitious than the course required,
and took quite a while to complete.<br /><br />
For the drones frame I finally had an excuse to purchase a 3d printer. I
modelled the frame and arms modularly to support future upgrades and
replacements from damage, reducing the cost of operation. Initially I
was using some 3d printed torodial propellers due to their higher
efficiency and lower sounds usage. Due to safety concerns of the
propellers not withstanding tension at higher RPMs and exploding I
switched to some generic acrylic propellers I bought. A challenge in
designing the frame was leaving enough room to contain the volume of the
wires. I went with the Elegoo Neptune 3 as the printer was open-source
and had a much better cost-to-utility than proprietary printers such as
the Ender series. I am not sponsored I just really like the printer.<br
/><br />
The flight computer was the most difficult part to program. Using an
ultrasonic distance sensor, gyroscope, and accelerometer the drone has
enough information to probably never crash. The autopilot is implemented
on an Arduino UNO using a
<a
href="https://en.wikipedia.org/wiki/Proportional%E2%80%93integral%E2%80%93derivative_controller"
>PID controller</a
>
for stabolization. The autopilot quality is currently impacted every time
the drones mass distribution changes, which can be fixed with a reinforcement
machine learning algorithm. For the higher level programming such as flight
automation, video transmission and WiFi communications I used a Raspberry
Pi Pico. There is also a 2.4GHz line of sight receiver for manual control.
A future upgrade may contain a SIM card for near-infinite remote control
connection, but drone regulations would make this difficult.<br /><br />
The power system is the most physically challenging portion of the
drone. The motors took 14.6 Volts, while the UNO microcontroller took 5
Volts, and the Pico and most sensors run at 3.3 Volts. All of the power
to the motors ran through the power distribution board, which I modified
to also output the lower voltages and used Bidirectional Logic Level
Converters to shift between 3.3 and 5V signals where needed. A potential
flaw with having all the power coming from the same source is spikes in
energy consumption to the motors may cause the microcomputers to receive
too little power, which could be fixed with a capacitor. Luckily, I
haven't experienced this yet as the 2C discharge rate on the 2.5Ah
capacity battery is more than enough. Having a battery this big does
mean it takes up about half the internal electronics volume and is half
of the drones mass, but it can also maintain full throttle for half an
hour making for long flights.<br /><br />
The motors I used were a bit overkill for a 1.1kg drone, as going past
20% throttle sends it shooting through the sky -- which is not a bad
issue to have. Here's the technical numbers behind that: I have
propellers with a 6cm radius on motors with a 2450KV rating (2450 rpm
per volt) at peaking at 14.6 Volts. From this the tip speed is computed
to be ~225m/s under no load at max throttle, quite a lot more than what
is safe or necessary to get into the air.
</p>

58
static/posts/gamedev/work-history.md
View File

@@ -1,58 +0,0 @@
<h1>Work Experience</h1>
<ul>
<li>
<h3>Cybersecurity Analyst for Equifax</h3>
<ul>
<li>Analyzed Security Posture, Business Risk, Compliance</li>
<li>
Supported Vulnerability Management sourced from Penetration
Tests & automated tools
</li>
<li>Handled issues, exceptions, and deviations</li>
<li>
Provided Security Advisment Services for infrastructure and
code development
</li>
<li>
Used automated scanning tools, GCP, AWS, shell scripts,
Linux
</li>
</ul>
</li>
<li>
<h3>Software Developer for Parkside Pool</h3>
<p>
This is a role that didn't exist before me. I was a lifeguard
and swim instructor here, and the pool I worked at was being
swamped with administrative work. Eventually I took on this work
for the money that came with it. For ten hours each week, I
would be reading emails, co-ordinating lessons, and sending out
schedules. Creating the schedules was a tedious task, with
siblings at different levels needing back to back lessons,
people not available some days or times, and other various soft
and hard requirements. <br /><br /> Soon I found out there is no
easy algorithm known for finding solutions to problems with many
soft and hard requirements. This is a NP-hard and is described
in the
<a href="https://en.wikipedia.org/wiki/Nurse_scheduling_problem"
>Nurse Scheduling Problem</a
>. Fortunately there was a cap of 128 students I could fit into
classes per week and the final algorithm was around O(n³) for n
swimmers, so the runtime wasn't horrible. <br /><br />
Implementing the reception and confirmation of lessons was straightforward
enough with a Flask site using Stripe for payments, and Firebase
service for sending confirmation emails.
</p>
</li>
</ul>
</div>
<div class="align-right card">
<h1>Volunteering</h1>
<p>
I volunteer within the Computer Science, Mathematics, and Statistics
department at the University of Toronto as a Computer Science Ambassador
to help cultivate interest in the Mathematical Sciences in high school
students, as well as providing direction for first year students within
the programs.
</p>