-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
282 lines (265 loc) · 13.1 KB
/
index.html
File metadata and controls
282 lines (265 loc) · 13.1 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
<!DOCTYPE HTML>
<!--
Paradigm Shift by HTML5 UP
html5up.net | @ajlkn
Free for personal and commercial use under the CCA 3.0 license (html5up.net/license)
-->
<html>
<head>
<title>Mark's Site</title>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=no" />
<meta name="description" content="" />
<meta name="keywords" content="" />
<link rel="stylesheet" href="assets/css/main.css" />
<link rel="icon" href="images/icon.png">
</head>
<body class="is-preload">
<!-- Wrapper -->
<div id="wrapper">
<!-- Intro -->
<section class="intro">
<header>
<h1>Mark Beshara</h1>
<p>B.S. Computer Engineering, OSU M.S. Computer Science, OSU</p>
<p>Applying AI to complex problems</p>
<ul class="actions">
<li><a href="#first" class="arrow scrolly"><span class="label">Next</span></a></li>
</ul>
</header>
<div class="content">
<span class="image fill" data-position="center"><img src="images/pic01_2.jpg" alt="" /></span>
</div>
</section>
<!-- Section -->
<section id="first">
<header>
<h2>About Me</h2>
</header>
<div class="content">
<p>Hi, I'm <strong>Mark Beshara</strong> and I'm programmer specializing in artificial intelligence. My interests are in a variety
of AI domains but I'm primarily interested in reinforcement learning and natural language processing. In addition to my primary interests I am also knowledgeable in
classical machine learning and various deep learning techniques. I hope to apply AI to challenging problems that currently cannot be solved by traditional means or
can be vastly improved through new technology. To do so, I am always looking for opportunities to learn new skills and knowledge. Currently I am attempting to live up to this
goal by pursuing job opportunities in industry to gain practical experience in applied ML/DL as well as looking into the possibility of obtaining a PhD in the hopes of contributing to the advancement of the field.</p>
<span class="image main"><img src="images/pic04.jpg" alt="profile picture" /></span>
</div>
</section>
<!-- Section -->
<section>
<header>
<h2>Skills </h2>
</header>
<div class="content">
<table>
<tbody>
<tr>
<th><strong>General Skills</strong></th>
<th><strong>ML Tools</strong></th>
<th><strong>Languages</strong></th>
</tr>
<tr>
<td>
<ul style="list-style-type:none">
<li>Software Development</li>
<li>Algorithms</li>
<li>Git</li>
<li>REST</li>
<li>Linux/UNIX Environments</li>
<li>Jupyter Notebooks</li>
<li>Statistical Machine Learning</li>
<li>Neural Networks/ Deep Learning</li>
<li>Data Mining</li>
<li>Web Scrapping</li>
<li>Computer Vision</li>
<li>Natural Language Processing</li>
<li>Reinforcement Learning</li>
<li>Optimization</li>
</ul>
</td>
<td>
<ul style="list-style-type:none">
<li>Pytorch</li>
<li>huggingface</li>
<li>spacy</li>
<li>Sci-kit Learn</li>
<li>Numpy</li>
<li>Pandas</li>
<li>Stable-Baselines</li>
<li>Gym</li>
<li>matplotlib</li>
<li>MLFlow</li>
<li>AzureML</li>
</ul>
</td>
<td>
<ul style="list-style-type:none">
<li>Python</li>
<li>Java</li>
<li>C</li>
<li>HTML/CSS</li>
<li>MATLAB</li>
<li>SQL</li>
<li>Any new language in 1-2 weeks</li>
</ul>
</td>
</tr>
</tbody>
</table>
</div>
</section>
<!-- Section -->
<section id="projects">
<header>
<h2>Projects</h2>
</header>
<div class="content">
<h3 style="color: black;">Table of Contents</h3>
<ul>
<li><a href="#otto" style="color: black;">OTTO - Multi-Objective Recommender System</a></li>
<li><a href="#idc9" style="color: black;">Visualizing ICD-9 Code Predictions</a></li>
<li><a href="#lisp" style="color: black;">Lisp Interpreter</a></li>
</ul>
<!-- Section -->
<!-- <section id="bitcoin">
<header class="projecthead">
<h3>RL Bitcoin Trader</h3>
<p> Taught a RL agent to trade Bitcoin with positive profit on unseen test data using historical crypto data and Stable-Baselines RL library.
</p>
</header>
<div class="content">
<p>This project is currently in development, but preliminary results will be posted soon.</p>
<p>Potential employers feel free to reach out to me for this projects code while I continue to work on a write up.</p>
</div>
</section> -->
<!-- <section id="disaster">
<header class="projecthead">
<h3>Kaggle Real or Not? NLP with Disaster Tweets Competition</h3>
<p> Project designed to classify tweets depending on whether or not the tweet represents discusses
a tragedy or not.
</p>
</header>
<div class="content">
<p>This project was done as part of the <a style="color: black;" href="https://www.kaggle.com/c/nlp-getting-started"> Real or Not? NLP with Disaster Tweets</a>
Kaggle Competition. The data provided from Competition was first processed, by tokenizing the sentences, and embedding each token using pre-trained GloVe word
embeddings. An average of all token embeddings was taken to create a feature vector that could be used in any machine learning classifier. A k-nn classifier
was then used to perform the classification, achieving 75% accuracy on the test set. An example repository can be found <a href="https://github.com/mark-beshara/DisasterTweetPrediction" style="color: black;">here</a>,
however please note this github repository contains an older version of the project that used random forest classifiers and attempted to optimize the hyperparameters using a grid search.
</p>
</div>
</section> -->
<!-- Section -->
<section id="otto">
<header class="projecthead">
<h3>OTTO Multi-Objective Recommender System</h3>
<p> Kaggle competition project which used transformers to recommend products to users based on their preferences.</p>
</header>
<div class="content">
<p>This Kaggle contest had the goal of predicting the next set of user actions as well as the product ids associated with those actions from any given point in a users action history. Given
the formulation I assumed that a good model would be a transformer since this could be treated as a generative task where the tokens are the product ids/user actions taken. Currently
I am trying to do exactly that using huggingface and attempting to train a GPT-2 model and a custom tokenizer on the given training data using this architecture.
</p>
<p>This project is currently in development, but preliminary results will be posted soon.</p>
</div>
</section>
<section id="idc9">
<header class="projecthead">
<h3>Visualizing ICD-9 Code Predictions</h3>
<p> Explainable AI project which visualized components of convolutional neural networks to better understand classification decisions.</p>
</header>
<div class="content">
<p>
This project focused on visualizing the attention scores from pre-trained CNN models for the multi-class classification problem of assigning ICD-9 codes to discharge summaries.
The project used the MIMIC-III dataset, and the pre-trained model was the CAML model, where one can read about <a style="color: black;" href="https://arxiv.org/pdf/1802.05695.pdf">here</a>,
and find the github repository <a style="color: black;" href="https://github.com/jamesmullenbach/caml-mimic">here</a>.
</p>
<p> Our project modified the original scripts to save the attention weights and outputs from the pre-trained model for a series of discharge summaries with true labels known. We visualized
the results by writing an interactive tool with D3 to visualize the attention score for each word in the discharge summary for each label. Unfortunately the MIMIC-III dataset cannot be
publicly released, so the D3 script is useless without it. However below you will find 2 presentations on this project with a more thorough explanation of the project as well
as some examples.
</p>
<p style="text-align: center;" ><iframe src="https://docs.google.com/presentation/d/e/2PACX-1vRAjC-vJWWqE_FWVjjHoT5Kdq_PC7knwtWoI0M61kTHxqRJWy0v396_YikIWZ_u5Syheq_hJ1NkVpFu/embed?start=false&loop=true&delayms=3000" frameborder="0" width="480" height="299" allowfullscreen="true" mozallowfullscreen="true" webkitallowfullscreen="true"></iframe></p>
<p style="text-align: center;" ><iframe src="https://docs.google.com/presentation/d/e/2PACX-1vQTLtY2VMnpvl_bTaq5KdeymVVOSopv_4_-YOcXX3pJ9sqr9JbYe7cis8n8nInmfqwxczZA8j7y__Or/embed?start=false&loop=true&delayms=3000" frameborder="0" width="480" height="299" allowfullscreen="true" mozallowfullscreen="true" webkitallowfullscreen="true"></iframe></p>
</div>
</section>
<section id="lisp">
<header class="projecthead">
<h3>LISP Interpreter</h3>
<p>A LISP Interpreter written in python that reads in LISP expressions from a file and evaluates them.
</p>
</header>
<div class="content">
<p> Unfortunately this project was part of a class project and therefore I cannon't publish the code publicly. However, an example input file and an example output to the console can
be seen below. Potential employers feel free to reach out to me and I am happy to share the source code privately.
</p>
<div class="gallery">
<a href="images/gallery/fulls/lisp_2.jpg" class="landscape"><img src="images/gallery/thumbs/lisp_2.jpg" alt="" /></a>
</div>
<p style="text-align: center;">Input file</p>
<div class="gallery">
<a href="images/gallery/fulls/lisp_1.jpg" class="landscape"><img src="images/gallery/thumbs/lisp_1.jpg" alt="" /></a>
</div>
<p style="text-align: center;">Console output</p>
</div>
</section>
</div>
</section>
<!-- Section -->
<section>
<header>
<h2>Get in touch</h2>
</header>
<div class="content">
<p>Let's make something <strong>interesting</strong>!</p>
<form action="https://formspree.io/meqrvvrl" method="POST">
<div class="fields">
<div class="field half">
<input type="text" name="name" id="name" placeholder="Name" />
</div>
<div class="field half">
<input type="email" name="_replyto" id="email" placeholder="Email" />
</div>
<div class="field">
<textarea name="message" id="message" placeholder="Message" rows="7"></textarea>
</div>
<div class="field half">
<input type="text" name="_gotcha" style="display:none" />
</div>
<input type="hidden" name="_subject" value="Website contact" />
<input type="hidden" name="_next" value="//mywebsite.com/thanks.html" />
</div>
<ul class="actions">
<li><input type="submit" value="Send Message" class="button primary"/></li>
</ul>
</form>
</div>
<footer>
<ul class="items">
<li>
<h3>Email</h3>
<a href="markapbeshara@gmail.com">markapbeshara@gmail.com</a>
</li>
<li>
<h3>Elsewhere</h3>
<ul class="icons">
<li><a href="https://www.linkedin.com/in/mark-beshara-1a4848104/" class="icon brands fa-linkedin-in"><span class="label">LinkedIn</span></a></li>
<li><a href="https://github.com/mark-beshara" class="icon brands fa-github"><span class="label">GitHub</span></a></li>
<li><a href="https://www.hackerrank.com/thegreatmarker" class="icon brands fa-hackerrank"><span class="label">HackerRank</span></a></li>
<li><a href="https://medium.com/@markbeshara" class="icon brands fa-medium"><span class="label">Medium</span></a></li>
</ul>
</li>
</ul>
</footer>
</section>
<!-- Copyright -->
<div class="copyright">© Untitled. All rights reserved. Design: <a href="https://html5up.net">HTML5 UP</a>.</div>
</div>
<!-- Scripts -->
<script src="assets/js/jquery.min.js"></script>
<script src="assets/js/jquery.scrolly.min.js"></script>
<script src="assets/js/browser.min.js"></script>
<script src="assets/js/breakpoints.min.js"></script>
<script src="assets/js/util.js"></script>
<script src="assets/js/main.js"></script>
</body>
</html>