Ryan Le

Finished a 9 month Software Engineering Bootcamp with Springboard! PERN full stack web development skills achieved. I'm now transitioning into Headstarter Software Engineering fellowship program where I will be building AI driven projects with other teammates. My gameplan: 1. Continue the data structure/algorithm grind. Daily LeetCode problems alternating between Python and JavaScript. Can try more languages as I acquire them. Learn at least one new algorithm or concept daily. 2. Continue acquiring breath/depth of knowledge in available technologies. There's a lot of libraries/tools out there so I want to focus on the most impactful and efficient ones with thorough documentation. 3. Develop sight-reading skills for code. Similar to sight-reading sheet music, I want to be able to look at any code and figure out how things work as quickly as possible. 4. Implement personal projects. Some ideas include abstracting Optometry decision making into a weighted graph data structure with the end node being the most accurate diagnosis. Subjective refraction can also be abstracted as well since input is simply a Boolean provided by the patient (one or two becomes 0 or 1). A directed graph would work nicely. 5. Find a paid internship/job. This would help apply my skills to a real world application and within a company framework. Also, I got bills to pay like everyone else. Very excited for what's to come! https://lnkd.in/gJUk-RkD

63

3 Comments

End of my AI Residency at Meta Today marked the end of my AI residency program. The last 2 years have been an incredible learning experience for me. I had the opportunity to delve into applied research while working on some amazing projects with some fantastic mentors and peers. I began my residency working on Horizon world understanding and asset understanding. I worked on developing image collection algorithms from VR and on vision modules training & recommendation problems. My first prod-release was the "For your world" section on the Horizon App (recommending similar assets for creators in VR worlds). After a reorg to GenAI in February of last year, I had the chance to work on AI characters. It was my first pure NLP project, coming from a vision/audio background. I was fortunate enough to witness the development of the project from a very ambiguous idea to a fully matured product. I worked on collecting open-source conversation datasets, experimenting with prompt templates & llm-finetuning, automated evals & synthetic data for alignment, automated prompt tuning & finally witnessed the launch at Meta-Connect last year (saw Zuck in person and got a picture clicked with him). After Meta-Connect, I worked on building models for prompt-suggestion & modeling human behavior in user-ai conversations. During the same time, I got a chance to work a bit on the AI creation side, and on alignment for avoid-topics. Finally, towards the end, I got the chance to witness the launch of full-fledged AI Studio. I had the privilege of working in one of the most fast-paced environments. I truly experienced the "Move Fast, break things" mantra. I had the opportunity to work in highly ambiguous scenarios with changing priorities and to come up with more directed goals. I got to explore applied research, plan projects from scratch, and work on some really cool ideas. I made many mistakes, learned tons of new things, met some really awesome people, and got to work on some really interesting projects. Big thanks to Yashar for bringing me to the team, being my continuous support for first 1.5 years, regular 1-1, continuously checking on me, invaluable career discussions & advice, getting me chance to work on some really cool projects & providing continuous support throughout first 1.5 years. Thanks to my mentors/peers Aasish, Asish, Bo, Anchit, and many more I couldn't list here for continuous support throughout 2 years, numerous valuable feedbacks, helping me set directions over ambiguous & ambitious goals and numerous advices to help me grow in my career. Thanks to Dhruv [Trust me he’s real life Sheldon Cooper] for being an awesome friend, good roommate, being with me on highs & lows, taking mock interviews & bear with me for all the craps I talk about. What’s next? Stay Tuned!

1,289

28 Comments

🚀 BostonHacks 2024! Last weekend, Andrew Sun and I won 2nd place for the CometCare track at BostonHacks 2024, which was our first Hackathon. We had a lot of fun working from the ground up to create a final product called GoodDeeds, a platform designed for NGOs, non-profits, and other local organizations looking to post about events requiring community involvent across the spectrum of urgency, ranging from severe post-disaster recovery efforts to more relaxed community service initiatives. 🏢 For Organizations Organizations can sign up to create events, specifying a location, urgency rating, volunteer needs, and a description of the event. 🙋 For Volunteers Individuals like you and me can sign up as volunteers with a specified “home” location and a distance radius we’re willing to travel. Volunteers can view all events within their set radius, filter by date and start time, and sign up. Their email is added to a mailing list for updates, and they can reach out to event organizers or ask questions to an OpenAI chatbot directly within the app. 🎮 XP Levels To encourage community involvement, we introduced a game-like feature: XP levels for individual users that increase with volunteering attendance. Higher XP points are awarded for events with higher severity ratings (like post-hurricane cleanup). Users and organizers can see a leaderboard displaying top volunteers with their XP points and approximate location. 📣 Real-Time Notifications We integrated a notification system, allowing organizers to alert individuals within a set radius about upcoming events or updates for events they’ve signed up for, powered by the Twilio SendGrid API. The emails look nice, too! 💻 Tech Stack We had a lot of fun learning Streamlit, a frontend library built in Python, along with building a Flask backend and using a PostgreSQL database hosted on Neon. The frontend was deployed on Streamlit Cloud, while the backend was deployed on Render. Given the short time constraint, we’re thrilled with what we accomplished, and we look forward to future development! 🌐 Check It Out! Demo Video: https://lnkd.in/ecJyGufP Live Site: https://lnkd.in/ey5ht6yG (starts cold, may take a while on first signup) Source Code: https://lnkd.in/ew7PJySZ

37

14 Comments

🎉 Excited to share that two of my recent projects have been accepted to major AI conferences! 🙌 Huge thanks to my amazing collaborators ryu jegwang, Wonpyo Park, Woohyun Han, Kyuyeun Kim, Namhoon Lee, Jaeho Lee from #Google and #POSTECH! 1️⃣ MaskedKD has been accepted to #ECCV2024! This work introduces a method for reducing the supervision cost in distilling ViTs. By masking input tokens based on student attention scores, we can save up to 50% of teacher FLOPs while maintaining student accuracy. [Preprint: https://lnkd.in/gUJn4D3a] 2️⃣ CushionCache, a project from my Google internship, has been accepted to #EMNLP2024 main track! We introduce a novel method to mitigate activation outliers in LLMs, significantly improving activation quantization and enhancing generation quality. Our approach discovers a prefix (the CushionCache) that regularizes activations and reduces quantization errors. [Preprint: https://lnkd.in/gtArpEF2]

71

4 Comments

🌟 𝐃𝐫𝐞𝐚𝐦𝐢𝐧𝐠 𝐁𝐢𝐠, 𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐁𝐢𝐠𝐠𝐞𝐫 🌟 Last May, I embarked on an incredible journey with WisdomAI right after graduating from UIUC, and it has been nothing short of a dream run! Starting my career with a passionate, dynamic team like this one has been an amazing experience. 🚀 Here’s what I’ve absolutely loved about my time at WisdomAI so far: 1️⃣ Ownership Sparks Passion When you’re empowered with the right motivation and a sense of ownership, the task transforms into a pursuit. I’ve cherished the freedom to take charge of my projects, driven by curiosity for achieving results rather than just deadlines. 2️⃣ Culture of Growth & Impact I’m deeply grateful to the founders and my team for cultivating an environment of insane growth while delivering great value to our customers every single day. Huge thanks to the founders - Soham, Sharvanath, Kapil, and Guilherme for fostering this environment. 3️⃣ The Power of Team Energy The motivation and drive of the people around you can make all the difference. Being surrounded by such an energetic, supportive team has truly set the trajectory for my career. As we step into this new year, we’re doubling down with more energy, sharper focus, and even bigger goals for our startup. Here’s to a year of innovation, impact, and scaling new heights together! 💡🌍 #StartupJourney #GrowthMindset #WisdomAI #Teamwork #Innovation #NewYearGoals #CareerJourney

136

7 Comments

🚀 Excited to share some insights from a recent paper titled "TourLLM: Enhancing LLMs with Tourism Knowledge" by Qikai Wei, Mingzhi Yang, Jinqiang Wang, Wenwei Mao, Jiabo Xu, and Huansheng Ning! 🌟 Highlights: Addressing the Gap: Large language models (LLMs) like ChatGPT and Llama excel in various NLP tasks but struggle with tourism-specific knowledge. To bridge this gap, the authors introduced a specialized dataset and model fine-tuning approach. Cultour Dataset: The team developed a supervised fine-tuning dataset called Cultour, tailored for the culture and tourism domain. This dataset comprises: Tourism knowledge base QA data Travelogues data Diverse QA data Introducing TourLLM: TourLLM is a Qwen-based model fine-tuned with the Cultour dataset. It aims to improve the quality of information provided about attractions and travel planning. Evaluation: The effectiveness of TourLLM was validated using both automatic and human evaluation methods. The researchers proposed a novel human evaluation criterion named CRA (Consistency, Readability, Availability). Results: The experimental results demonstrated that TourLLM significantly enhances the quality and accuracy of responses in the tourism domain, offering more reliable travel advice and enriching user experience. 🔍 Key Takeaways: Customized Fine-Tuning: By creating domain-specific datasets and fine-tuning models accordingly, we can significantly improve the performance of LLMs in specialized areas. Innovative Evaluation Metrics: Developing tailored evaluation criteria, like CRA, ensures that the models meet the specific needs and expectations of users in different domains. Research Paper:https://lnkd.in/gruBhP6T #AI #MachineLearning #Tourism #NLP #LLMs #DeepLearning #Innovation

1

Week 4 of CUNY Tech Prep: I'm excited to be working with Daniel Chen and Andrew Lam, RT(R) throughout this semester in building a simple reinforcement learning data science project. Daniel is amazing at coding and has won several Hackathons in the past, including the CTP Hackathon that I participated over this past summer. Andrew and I met through a mutual friend, and we have many interests in common, including a background working in healthcare. This week, I watched a video on Clustering in Data Science, which I have attached below for your reference. This video introduced me to the concept of Clustering, which is the grouping of data based off a measure of similarity. The video continues to show a demo of creating dendrograms from a dataset in R. I thought that it was interesting because I recently worked with R and RStudio in a research project for Collegiate Science and Technology Entry Program (CSTEP). I am excited to recreate the same demonstration using Pandas and NumPy in Python.

7

Ever wondered why LLMs struggle with math? In this piece, I take a look under the hood of LLMs and math-related AI study tools, to uncover the reasons LLMs are not well-suited for math, what innovators are doing to get ahead of the problem, and some exciting companies in the space. I explore several approaches to improve LLM math performance: process supervision, integration with symbolic computing systems (i.e. ChatGPT + Wolfram|Alpha LLC), and novel ways of capturing unique and structured data ( Thinkverse (formerly Learnie AI), Photomath, Thetawise, among many others). One startup in Korea — QANDA(Mathpresso) — has built its own math-specific LLM that can outperform many of the incumbents. If you are building AI to promote math learning and understanding, please reach out! Very grateful for the opportunity to go deeper in the EdTech space at Reach Capital this summer before kicking off my MBA/MA in Education at Stanford in the fall! Shoutout to Tony Wan and James Kim for the support on this piece! https://lnkd.in/g8xk_hqJ

50

Exciting news from Cornell University! 🌟 Cornell has joined the U.S. Commerce Department's AI Safety Institute Consortium (AISIC) to support the development and deployment of trustworthy and safe artificial intelligence technologies. As a Computer Science freshman at Cornell's College of Engineering, this initiative resonates deeply with my aspirations in AI and machine learning. 🤖 The AISIC consortium, established by the National Institute of Standards and Technology (NIST), aims to bring together AI creators and users, academics, nonprofit organizations, and government and industry researchers. With more than 200 members, including companies and startups of all sizes, research teams, and representatives of professions deeply engaged in AI's use, the consortium is set to develop evidence-based guidelines and standards for AI measurement and policy. 🌍 Cornell's involvement in the consortium underscores its commitment to advancing safe and trustworthy methods, practices, and policies in AI, contributing to positive global impact. Across its campuses in Ithaca and New York City, Cornell is a hub of AI developers and users, including researchers who are creating new ways to use knowledge, ideas, and tools to have a positive global impact. 🌟 I am proud to be part of a university that is at the forefront of addressing the critical need for safe and trustworthy AI. This initiative aligns perfectly with my goal of becoming a Machine Learning engineer and exploring how AI can enhance productivity and competitiveness in business. 📈 #AI #MachineLearning #CornellUniversity #AISIC #AIethics #TechInnovation https://lnkd.in/e9rYfbtf

10

My time as a research engineering intern at Google DeepMind has come to a close! As someone whose world revolves around networked, computer, and distributed systems, I never imagined I'd land on my dream team – the model partitioning team! This summer, I worked on PartIR's automatic model partitioning capabilities(https://lnkd.in/egp5Ygvy), diving deeper into distributed computing and exploring applications in reinforcement learning. It was definitely a challenging project, but I picked up so many invaluable skills, from software engineering to research in an industry setting. I’d like to give a huge thank you to Sami Alabed, Masha Samsikova, and the entire model partitioning team for making me feel so welcome and helping me successfully complete my project! I’m also thrilled that my final presentation was well received (and even got some laughs!), helping many at GDM understand a complex topic. Outside of my project, I had the chance to meet so many fascinating people across different teams and, of course, the other interns! A big thanks to the GDM Academy organizers, Charlotte Colson and Claudia Pope, for their incredible support in running such a fantastic program while I was balancing my thesis. And to Lucia Lopez Rivilla and Ioana Mihăilescu, former GDM Academy interns, for helping me find my way around the company! This summer wasn’t without its challenges, though. I finally addressed the anxiety issues I’d been struggling with since my teenage years, as they were impacting my studies and relationships. Unfortunately, I wasn’t able to attend my final year exams as I had hoped. However, I'm so grateful to UCL's Computer Science team and Faculty of Engineering for giving me the flexibility to extend my studies by a year, without penalty, and extending my scholarship. I couldn’t be more thankful for their support! All in all, this experience has solidified my determination to pursue research. In my third year, I discovered my passion for networked systems, and during my time at AMD, I knew I wanted to work on distributed systems. Now, after my time at GDM, I’m confident that distributed and networked systems are my future – and with the added experience of AI, I feel more prepared to navigate the machine learning-filled computer science landscape. Here’s to a (hopefully) restful but equally exciting year ahead! See you soon, Morgane 🦆

219

16 Comments

Proud to share our work on enhancing LLM efficiency through dynamic early exit strategies! Collaborating with Juan D. Castano and Amir Voloshin, we built on Meta’s LayerSkip framework to introduce interpretable heuristics that identify when token predictions stabilize. Excited about the potential this unlocks for faster, more efficient LLM inference. A big shoutout to my amazing teammates for their hard work and insights!

22

CUNY Tech Prep: This week, we explored open-source models with Hugging Face. For my wordle group project, I’m planning to incorporate Meta's NLLB (No Language Left Behind) translation model, which supports over 200 languages. The model uses Torch tensors, essentially multi-dimensional arrays that efficiently process large-scale data, to handle translations smoothly. By integrating NLLB, I can offer translations and personalized hints in real-time, making the game accessible to a diverse audience. Checkout Meta's model here: https://lnkd.in/eCbZJz6x #MachineLearning #DataScience #CTP #HuggingFace #Meta

27

Finalist in the HeadStarter SWE Fellowship Hackathon🏆 Over this weekend, my team and I participated in the Headstarter SWE Fellowship hackathon, and I am happy to share that we were selected as one of the finalists for the Retell AI track. 💻 About our project: Our project, SpeakStudy, aims to maximize study efficiency while minimizing the time spent. Students can simply upload their lecture notes, and within seconds, SpeakStudy will assist you in your understanding. Students can use the microphone to express their doubts related to those uploaded lecture notes, and SpeakStudy will respond by clarifying those doubts with voice technology, creating a productive and insightful conversation for the student. 📚 Key Takeaways: 🌱 Stepping out of my comfort zone: Participating in the HeadStarter SWE Fellowship Hackathon was a significant step outside of my usual routine. I usually take my time when making projects; however, hackathons are fast-paced, demanding environments where innovation and quick thinking are crucial. On top of that, I was working with a tech stack that I had minimal experience with. Therefore, I embraced the technical and strategic challenges, pushing the boundaries of what I thought I was capable of. In the end, this experience has not only expanded my skill set but also boosted my confidence in tackling unfamiliar challenges. 🤝 Teamwork and collaboration: My team consisted of individuals with diverse skill sets, and our success as finalists was a direct result of our ability to communicate and integrate our strengths. Right from the start, we drew the architecture of our product and distributed our roles based on our strengths and interests. We quickly established a smooth workflow and supported each other under tight deadlines. This experience reinforced the value of working in a collaborative environment where each member's input is valued. ⏱️ Problem-solving under pressure: This hackathon’s 36-hour timeframe created an intense atmosphere where problem-solving had to be quick and efficient. We encountered several technical and conceptual challenges, from integrating voice technology with the application to ensuring that the AI could accurately respond to a wide range of student queries. These challenges had to be resolved rapidly, often requiring us to think on our feet and adapt our approach along the way. This experience sharpened my ability to remain calm under pressure, think critically, and deliver solutions in a high-stakes environment. Thank you, Retell AI and Headstarter, for giving us this opportunity. I am also grateful to Yasin Ehsan 🚀, Faizan Ahmed, and Sajjaad Khader for organizing this event! Lastly, thanks to the team behind this product: Cameron Armijo, Shakib Absar, Natalia Rumbuc. We will continue to develop the website more and hopefully attract more users in the future. See the GitHub link below: https://lnkd.in/dPtCUTWF

40

5 Comments

What feels better than finishing a project? Seeing that your software is still being used after two years! In August 2022, after many pandemic delays, I finally made it to Japan for my research stay at the University of Tokyo. I was excited to join an experiment testing the Standard Model, as understanding this theory was one of my main motivations for studying physics. The experiment was brilliantly engineered, but the repetitive tasks to control it made it difficult to run and scale. This slowed our experiment down. So together with Nagase-san, who just started his PhD in the group, we decided to change this: 💻 Automated repetitive tasks with Python, turning manual tasks into reusable code. 🌱 Made the experiment re-runnable and scalable. ⏳ Saved tens of hours each week with automation. This shift allowed us to run the experiment more efficiently and get new results. The code has become an essential part of Nagase-san's daily work as he finishes his PhD thesis, and it always makes my day when I see his commits. What is your experience with writing software for data science or research? Let me know what tools help you the most!

28

9 Comments

Here’s how DoorDash evaluates its RAG performance by using LLM As A Judge. It's able to monitor retrieval correctness, response accuracy, coherence, and relevance to support questions which are posed by Dashers. 1. First, thousands of chat transcripts from Dashers are reviewed to identify common issues. 2. The LLM Judge is prompted with open-ended questions to assess them, and its answers are summarised into actionable feedback. 3. From these, high-frequency issues are built into prompts or rule-based metrics for continuous monitoring. 4. Human review teams validate the LLM’s evaluations by reviewing random transcript samples. This human feedback is calibrated with the automated system for better coverage. LLM Judge helps address issues like inaccurate retrieval, hallucinations, and low quality prompts. It makes sure that there's a continuous quality improvement pipeline alongside human support agents. Link to their full report is in the comments. #AI #LLMs #RAG

90

3 Comments

Here's an email from October that I never thought I'd write: asking a prof for an extension to help with juggling school, a project in Boston, a project in SF, and interviews--all in one week. I spent four days at Harvard for a hackathon, landed back at college on Monday to tackle school work, had two interviews on Wednesday and Thursday, and left Thursday for a hackathon at Berkeley. It was overwhelming, and I realized that you really can't do it all. You have to prioritize (and thus also deprioritize). Here’s how I approached it: 1. Interviews (I need a job lol) 2. Hackathons (I wanna win 🤩) 3. College (apparently, I need a degree, lol) Having too much to do--and being unable to do it all perfectly--is honestly what I find exciting about life. It keeps you on your toes. And it forces you to continually learn and grow.

401

38 Comments

🌟 Translate4Good, at Hack for Social Impact 🌟 Tackling the problem statement given by UN Convention to Combat Desertification(UNCCD), I collaborated with a talented team Alma Gashi, Nikita Damle and Alice Qingyu Z. to build an AI-powered translation system using Fetch.ai's autonomous agents. Our solution helps UNCCD translate complex technical content across six UN languages. Key achievements: 🚀 🚀 - Created efficient workflow and intuitive UX for professional UN translators to handle context-specific translations. - 93% Semantic Similarity score in English-Spanish translations, after fine-tuning MarianMT model with UN Parallel Corpora. - Built specialized Fetch.ai agents for document classification, terminology extraction, and quality validation. - Integrated adaptive learning from translator feedback. - Contributing to make UN documentation more accessible while exploring the potential of autonomous AI agents! - Won 2nd Place under Fetch.ai track🥈 Special thanks to our judging panel - Sana Wajid, Preston Bowman, Mark Losey, Sai Mounika Peteti and Fetch.ai team for recognizing the potential of our project. Thanks to Shelly C. Cher Hu Steven Echtman for putting together another impactful event. Huge shoutout to Fetch.ai Innovation Lab, United Nations and Microsoft for Startups for fostering a culture of innovation and enabling tech projects that drive social impact. #AI #Hackathon #FetchAI #UnitedNations #TechForGood #SocialImpactHack #BuildFetch

28

5 Comments