-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
195 lines (173 loc) · 12.4 KB
/
index.html
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
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Faraz Khoubsirat - Software Engineering Student</title>
<link rel="stylesheet" href="stylesheets/styles.css">
<!-- Add OpenGraph meta tags for link preview image -->
<meta property="og:image" content="jarvis.jpg">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="630">
</head>
<body>
<div class="wrapper">
<header>
<img src="profile.png" alt="Faraz Khoubsirat" class="profile-picture" width="240" height="200">
<h1>Faraz Khoubsirat</h1>
<p>Software Engineering Student<br>University of Waterloo</p>
<p class="view">
<a href="https://github.com/farazkh80">GitHub</a> |
<a href="https://www.linkedin.com/in/farazkh80/">LinkedIn</a> |
<a href="https://twitter.com/FarazDoTAI">Twitter</a> |
<a href="https://medium.com/@farazkhoubsirat80">Medium</a>
</p>
</header>
<section>
<p>I am a fourth-year Software Engineering student at the University of Waterloo, specializing in Machine Learning and Artificial Intelligence.
My passion lies in training and optimizing large transformer-based multimodal models, with a focus on enhancing their efficiency,
interpretability, and real-world applicability.</p>
<h2>Interests</h2>
<ul>
<li>Transformer-based efficient model training and inference</li>
<li>Multimodal LLMs and Vision LLMs</li>
<li>Parameter efficient model finetuning</li>
<li>Transformer-based Embodied Autonomous Agents</li>
<li>Tech, investing, personal growth, podcasts, and books</li>
</ul>
<h2>Skills</h2>
<p><strong>Languages:</strong> Python, C++, C, SQL, JavaScript, Java, Scala, Assembly</p>
<p><strong>Technologies:</strong> PyTorch, Jax, TensorFlow, Pandas, Scikit-learn, MySQL, Docker, Kubernetes, GCP, AWS</p>
<p><strong>ML Focus:</strong> Modeling and Training Transformer-Based LLMs and Vision LLMs, Parameter-Efficient Fine-Tuning of LLMs, Optimized Export and Inference of LLMs, Retrieval-Augmented Generation</p>
<h2>Work Experience</h2>
<h3>Software Engineer Intern - NVIDIA</h3>
<p>Santa Clara, CA | Oct 2024 - Dec 2024</p>
<ul>
<li>Working with Deep Learning Frameworks Inference-TensorRT (TRT) team to develop Tripy, a new PyTorch-like Python machine learning framework on top of TRT.</li>
</ul>
<h3>Research Engineer Intern - Cohere AI</h3>
<p>Toronto, ON | Jan 2024 - Aug 2024</p>
<ul>
<li>Designed parameter-efficient fine-tuning framework for model-parallel distributed training and inference of LLMs on GPUs</li>
<li>Implemented, trained, evaluated, and deployed Cohere's first Multimodal Vision LLM in PyTorch</li>
<li>Led internal research to evaluate fine-tuning techniques for LLMs on customer data, comparing methods such as parameter efficient supervised fine-tuning, direct policy optimization, and offline preference tuning</li>
<li>Developed and dockerized a FastAPI + Uvicorn server for batch inference of fine-tuned LLMs and VLLMs, integrating it with evaluation libraries to enable large-scale evaluation of fine-tuned models</li>
</ul>
<h3>Machine Learning Engineering Intern - Cohere AI</h3>
<p>Toronto, ON | May 2023 - Present</p>
<ul>
<li>Led the development of a comprehensive data-quality evaluation library to evaluate LLM quality in terms of human preference, grammar, spelling, and repetitiveness, using Python and Pandas</li>
<li>Integrated the data-quality library to run on training datasets and +10,000 daily API data to flag and remove bad items</li>
<li>Fine-tuned Cohere's Command model on collected and cleaned API data resulting in a 70% performance boost, using PyTorch and Scikit-learn</li>
</ul>
<h3>Machine Learning Engineering Intern - Cohere AI</h3>
<p>Toronto, ON | Sep 2022 - Dec 2023</p>
<ul>
<li>Implemented and benchmarked throughput-efficient, LLM inference setups of Jax, PyTorch+TensorRT, and TensorFlow frameworks running on CPUs, GPUs, and TPUs</li>
<li>Leveraged mixed-precision inference, kernel fusion, data parallelism, and batch size tuning techniques to boost the inference latency of models on CPUs by 4X</li>
<li>Conducted a self-initiated, proof of concept project on inference of embedding models on TPUs using Jax and Haiku frameworks, resulting in 2X throughput improvement and 50% cost reduction compared to A100 GPUs</li>
<li>Built a dockerized end-2-end model export pipeline in Python for handling export and deployment of billion-parameter Jax model checkpoints, to TensorFlow, PyTorch and Triton FasterTransformer serving solutions for production</li>
</ul>
<h3>Machine Learning Engineering Intern - Airy3D</h3>
<p>Montreal, QC | Jan 2022 - Apr 2022</p>
<ul>
<li>Conducted data analysis, cleaning, validation, and feature engineering on 500k+ unstructured image data</li>
<li>Modeled, trained and hyperparameter tuned a custom end-to-end CNN built on top of Resnet in PyTorch, leveraging sensor generated depth maps to improve state-of-the-art object tracking F-Score by 12%+</li>
<li>Implemented a Gauss-Newton based optimizer with Conjugate Gradient in PyTorch and Python, improving object localization latency by 10X compared to a standard Gradient descent</li>
<li>Wrote CUDA kernels for model sub-modules and deployed the model as a desktop application via C++ and ONNX</li>
</ul>
<h3>Software Developer Intern - Ford Motor Company of Canada</h3>
<p>Oakville, ON | May 2021 - Aug 2021</p>
<ul>
<li>Developed automated test scripts using <strong>Python</strong> and <strong>Slash</strong> library, expanding the Wi-Fi testing coverage by <strong>10%</strong></li>
<li>Improved overall runtime and test pass rate by <strong>40%</strong> through debugging and fixing software of failed test suites</li>
<li>Executed <strong>1000+</strong> automated test runs via <strong>Jenkins</strong> CI, identified root cause of failures, and reported defects</li>
</ul>
<h3>Software Developer Intern - Vancouver Community Network</h3>
<p>Vancouver, BC | May 2020 - Aug 2020</p>
<ul>
<li>Redesigned client-side authentication system by implementing a SHA-512 hashing algorithm in <strong>C</strong> and <strong>Perl</strong>, minimizing system breaches and increasing identity validation speed by <strong>80%</strong></li>
<li>Reduced local server dependency by building a RESTful API in <strong>Python</strong> to migrate user data to GSuite</li>
<li>Increased team efficiency by <strong>+10</strong> hours a week through automating compromised password detection using <strong>Node.js</strong></li>
</ul>
<h2>Research Experience</h2>
<h3>Undergraduate Research Assistant - Stanford NLP</h3>
<p>Remote | April 2023 - Present</p>
<ul>
<li>Conducting research on DSPy framework and Retrieval-Augmented Generation (RAG) with LLMs (LLama3, GPT-4) under the supervision of Omar Khattab</li>
<li> <strong>Publication</strong>: Prompts as Auto-Optimized Training Hyperparameters: Training Best-in-Class IR Models from Scratch with 10 Gold Labels <a href="https://arxiv.org/abs/2406.11706">https://arxiv.org/abs/2406.11706</a></li>
</ul>
<h3>Undergraduate Research Assistant - University of Waterloo Data Systems Group</h3>
<p>Waterloo, ON | April 2023 - Present</p>
<ul>
<li>Conducting research on Retrieval-Augmented Generation (RAG) with LLMs (LLama3, GPT-4) using PyTorch and Huggingface under the supervision of Ronak Pradeep and Prof. Jimmy Lin</li>
</ul>
<h2>Publications</h2>
<ul>
<li>
<strong>Prompts as Auto-Optimized Training Hyperparameters: Training Best-in-Class IR Models from Scratch with 10 Gold Labels</strong>
<br>
<a href="https://arxiv.org/abs/2406.11706">https://arxiv.org/abs/2406.11706</a>
</li>
<!-- Add more research papers here if available -->
</ul>
<h2>Blog Posts</h2>
<ul>
<li>
<strong><a href="https://medium.com/@farazkhoubsirat80/hacking-images-with-llms-part-i-encoding-alignment-combining-e594f18491a4">Hacking Images with LLMs, Part I: Encoding, Alignment, Combining</a></strong>
<br>
Published on Sep 22, 2024 · 5 min read
<br>
<em>With the rise of open-source vision transformer models, image understanding and multimodal LLMs have become essential features in AI. In this blog, we'll explore how to get LLMs to process images...</em>
</li>
<li>
<strong><a href="https://medium.com/@farazkhoubsirat80/building-your-ml-career-essential-advice-for-students-and-practitioners-9b1bc70c558c">Building Your Machine Learning Career: Essential Advice for Students and Practitioners</a></strong>
<br>
Published on Aug 31, 2024 · 3 min read
<br>
<em>Over the past few years, countless students have asked me how to break into the world of machine learning. Since my first year of college…</em>
</li>
<li>
<strong><a href="https://medium.com/@farazkhoubsirat80/top-5-lessons-i-learned-from-analyzing-1000-yc-startups-5aa149d51894">Top 5 Lessons I Learned from Analyzing 1000 YC Startups</a></strong>
<br>
Published on Mar 30, 2024 · 6 min read
<br>
<em>As a person in tech, I often wondered what kind of companies get into Y Combinator, the Ivy League of startup accelerators.</em>
</li>
<li>
<strong><a href="https://medium.com/@farazkhoubsirat80/how-to-create-a-personal-website-with-notion-and-github-in-less-than-10-minutes-2f8ff0bd3e2f">How to Create a Personal Website with Notion and Github in less than 10 minutes</a></strong>
<br>
Published on Dec 26, 2022 · 2 min read
<br>
<em>Are you dreaming about making a quick personal website using the "elegant" notion formatting?</em>
</li>
</ul>
<h2>Must Read</h2>
<ul>
<li><a href="https://arxiv.org/abs/2203.02155">Training language models to follow instructions with human feedback</a></li>
<li><a href="https://arxiv.org/abs/2204.06514">Scalable Training of Language Models using JAX pjit and TPUv4</a></li>
<li><a href="https://jalammar.github.io/illustrated-stable-diffusion/">The Illustrated Stable Diffusion</a></li>
<li><a href="https://huggingface.co/blog/annotated-diffusion">The Annotated Stable Diffusion</a></li>
<li><a href="https://arxiv.org/abs/2210.03094">VIMA: General Robot Manipulation with Multimodal Prompts</a></li>
<li><a href="https://www.notion.so/Paper-Summaries-07ca1f2d4079446bb5fc84d530489293?pvs=21">More</a></li>
</ul>
<h2>Must See</h2>
<ul>
<li><a href="https://vimalabs.github.io/">VIMA</a></li>
<li><a href="https://lexica.art/">Lexica</a></li>
<li><a href="https://podcast.ai/">Joe Rogan Interviews Steve Jobs</a></li>
</ul>
<h2>Must Watch</h2>
<ul>
<li><a href="https://www.youtube.com/watch?v=Gfr50f6ZBvo&t=8s">Demmis Hassabis and Lex Fridman: DeepMind AI, Superintelligence & the Future of Humanity</a></li>
<li><a href="https://www.youtube.com/watch?v=cdiD-9MMpb0&t=5273s">Andrej Karpathy and Lex Fridman: Language Models, Tesla AI, Tesla Optimus</a></li>
<li><a href="https://www.youtube.com/watch?v=kFQUDCgMjRc&t=3576s">Chamath Palihapitiya and Lex Firdman: Money, Success, Startups, Energy, Happiness</a></li>
<li><a href="https://www.youtube.com/watch?v=3qHkcs3kG44">Naval Ravikant and Joe Rogan: Money, Happiness, Startups Purpose of Life</a></li>
</ul>
<h2>My Book Collection</h2>
<a href="https://www.notion.so/Reading-List-172d27193245440eaa60c6c0bbca3d85?pvs=21">Reading List</a>
</section>
</div>
<script src="javascripts/scale.fix.js"></script>
</body>
</html>