Weihua Zhu

Agentic AI is revolutionizing the industry of legacy code migration and modernization where we see clear positive business and societal impacts of this disruptive technology.

6

How to write a killer Software Engineering résumé? By Terrence Kuo An in-depth analysis of the résumé that got me interviews at Google, Facebook, Amazon, Microsoft, Apple, and more. This résumé got me interviews at Google, Facebook, Amazon, Microsoft, and Apple.I obtained these interviews by sending my résumé to the résumé black hole, also known as applying online. Applying online is the most common way people go about applying for a job and therefore the least effective way to land an interview due to competition. Yet that is exactly how I obtained all my interviews. How did I accomplish this? In this article, I will go through a line-by-line analysis of my résumé for the following purposes: explaining the choices that I made in creating my résumé why I believe this résumé worked to help me land those interviews, and how you can create an even better résumé! I decided to write this article because I struggled a lot with landing interviews when I first started looking for a job. It would have been extremely helpful for me to have a real-life example résumé to look at. This article is organized into the following sections: The All Too Familiar Way of Not Landing an Interview - a short anecdote of my frustrations when I first started applying for jobs Evaluating the Options: Moving Forward - a reflection on different strategies to improve the odds of landing interviews Learning How to Write a Killer Résumé By Example - the step-by-step analysis of my résumé with each of the following sections corresponding to my résumé: 1.The Essentials from a Glance 2.Who Are You 3.Contact Information 4.Education 5.Employment 6.Personal Projects 7.Skills

111

1 Comment

🚀 How Facebook Scaled Memcache to Handle Billions of Users 📈 I have been reading the book by Alex Xu called System Design Interview an Insider’s Guide, and he recommended reading the paper published by Facebook (Meta now :D) to understand how Facebook handled most of the common caching issues. In the world of high-scale systems, efficient caching is crucial for maintaining performance and responsiveness. My latest article on Medium dives deep into how Facebook optimized and scaled Memcached to handle the demands of billions of users. 🔍 What you’ll learn: - The role of consistent hashing in load balancing - Techniques to reduce latency and prevent cold cache misses - Strategies for failover and maintaining data consistency - How Facebook minimized network congestion and optimized traffic 💡 Scaling is not just about adding more hardware; it’s about building smart systems that adapt to changes and grow alongside the user base. Check out the full article here: https://lnkd.in/eXk4DQmd #Scaling #DistributedSystems #Caching #FacebookEngineering #TechLeadership #Memcached #SystemDesign #CloudComputing #HighPerformanceComputing #SoftwareEngineering

12

2 Comments

Meet Jordi A., a 4th-year PhD student in Computer Science at the University of Edinburgh. This summer, he joined Meta's Fundamental AI Research (FAIR) team as a Research Scientist Intern. "I'm working on large language models and code generation. I'm still astounded by the massive number of GPUs and the level of the research collaborators I have access to at Meta," Jordi explained. "I'm investigating how to get AI models to have a better understanding of code by using information obtained when executing programs." Jordi was drawn to Meta because of FAIR's groundbreaking open science and ambitious plans for Artificial General Intelligence (AGI). "I'm so excited to be part of a lab working on the frontiers of AI research," he said. The best advice he received this summer? "Make sure you meet all the relevant people — there's always something to learn from other's work. It's easier to limit yourself to your immediate peers but going the extra mile to meet researchers and interns from other teams or locations has been totally worth it." After completing his internship, Jordi plans to finish his PhD and continue researching AI for code generation. "There is so much more that AI can do for us — this is just the beginning.” #LifeAtMeta #AI #AIResearch #MetaIntern

19

1 Comment

🦙 The new Llama 3.1 models with the new NVIDIA NeMo Retriever NIMs create a powerful #generativeai duo. ➡️ Technical deep dive on how to build an agentic RAG pipeline by the newest NeMo Retriever embedding and reraking NIM microservices. https://bit.ly/3WlUSbM

1

🚀 Making LLMs work for you - A Guide to Become 10x Programmer/MLE/DS I am sure many of you use LLMs for your day-to-day work. Mostly, I have seen people use LLMs while writing code (Copilot), or as an alternative to Copilot to catch bugs, but they have become much more useful than that, provided you know how to use them well. Here are a few ways I have been using LLMs in my daily work that are helping me increase my productivity: 📚 1. Quick Paper Assessment with ELI5 Summaries - Paper abstracts are called abstracts for a reason. Asking an LLM for an ELI5 summary makes it much easier to know if I want to read a paper. 💻 2. Implementing Research Papers in Code - Try giving an LLM like Claude/ChatGPT a research paper and ask to code it. You would be surprised. Tip: Claude performs much better for this task. 🔄 3. Upgrading Code Structure and Readability - Do you want to add typing hints when writing your Python code, or do you need comments in your code, or do you have a big function that you want to be converted to classes in your code? LLMs can do it all, and you would be amazed at the code-conversion power they have. ⚡ 4. Optimizing Code Performance - The thing with LLMs is that if you tell them what to do, they can do it well. For example, if you give them a non-optimized piece of code that runs correctly, they will be able to write it in a much-optimized way. For example, I just used an LLM to convert a matrix multiplication in PyTorch to sparse MM as I was getting an OOM memory error because one of the matrices became too large. I am not even joking, but that would have taken at least 1-2 hours for me to come up with. 📝 5. Automated Documentation - Got bored of writing documentation for what you did but have the code? LLMs also work the other way around. Want to write a summary of what you did in a week, you can use an LLM for that. Saves so much time that gets wasted in formatting. 🧪 6. Streamlined Test Framework Creation - Let LLMs handle the boilerplate for Unit tests & Integration tests while you focus on core logic. 🌐 7. Language-Agnostic Development - You can paste a function in Python and ask for its equivalent in Rust, complete with language-specific best practices and error handling. This dramatically reduces the learning curve when working with new languages and helps maintain productivity during language transitions. 🤝 Apart from all of the above, LLMs work well as mentors as well as interns. LLMs are great when you just need someone to bounce ideas with. I sometimes just talk to an LLM as I would to a colleague talking about the problems I have and how I can solve them. And they come up with the right ideas or send me in the right direction a lot of times. They also act as a personal assistant for free, who can deal with writing messages, creating charts, and doing all sorts of uninteresting work for you. 💭 What's your favorite way to use LLMs? Let me know in the comments! #MachineLearning #AI #Programming

47

Efficient Location Indexing and Scalability in Uber's System Design Came across an interesting article by Neo Kim about how How Does Uber Find Nearby Drivers? Highlights: - Uber uses the H3 library for location indexing, which divides Earth's surface into cells on a flat grid for efficiency. - H3 supports efficient cell retrieval using latitude, longitude, and resolution parameters. - Uber stores raw locations using Apache Cassandra for durability and adds a Redis cache layer for read operations. - Map matching is done to transform noisy GPS signals into actual road segments. - Uber partitions services using ringpop for scalability and uses gossip protocol to track server status. Neo explained the concepts in an easy to read and understandable manner. You can also checkout similar articles in his newsletter "System Design Newsletter" Subscribe : https://lnkd.in/gxK92GgK Link to article : https://lnkd.in/g35SGyTe #systemdesign #uber

24

1 Comment

🔓 Unlock the power of your data with NVIDIA NeMo Curator. Discover how to #customize and optimize your model by integrating high-quality, domain-specific datasets. https://nvda.ws/3WO1QYZ

1

Chip Huyen on Twitter today below. I 100% agree. Metric hygiene is system hygiene. LLM-Generated metrics require a higher standard of care than they're currently receiving, and much more scrutiny. Mind the compasses. 🐦  A big issue I see with AI systems is that people aren't spending enough time evaluating their evaluation pipeline. 1. Most teams use more than one metrics (3-7 metrics in general) to evaluate their applications, which is a good practice. However, very few are measuring the correlation between these metrics.If two metrics are perfectly correlated, you probably don't need both of them. If two metrics strongly disagree with each other, either this reveals something important about your system, or your metrics just aren't trustworthy.2. Many (I estimate 60 - 70%?) use AI to evaluate AI responses, with common criteria being conciseness, relevance, coherence, faithfulness, etc. I find AI-as-a-judge very promising, and expect to see more of this approach in the future.AI-as-a-judge scores aren’t deterministic the way classification F1 scores or accuracy are. They depend on the judge's model, the judge's prompt, and the use case. Many AI judges are good, but many are bad.Yet, very few are doing experiments to evaluate their AI judges. Are good responses given better scores? How reproducible the scores are -- if you ask the judge twice, do you get the same score? Is the judge's prompt optimal? Some aren’t even aware of the prompts their applications are using, because they use prompts created by eval tools or by other teams.Also fun fact I learned from a (small) poll yesterday: some teams are spending more money on evaluating responses than on generating responses 🤯 https://lnkd.in/gYnTRZW3

2

🚗 𝐌𝐚𝐬𝐭𝐞𝐫𝐢𝐧𝐠 𝐋𝐨𝐰-𝐋𝐞𝐯𝐞𝐥 𝐃𝐞𝐬𝐢𝐠𝐧: 𝐓𝐡𝐞 𝐔𝐛𝐞𝐫 𝐂𝐚𝐬𝐞 𝐒𝐭𝐮𝐝𝐲 In tech interviews, particularly for backend roles, understanding the Low-Level Design (LLD) of complex systems like Uber can set you apart. Here's a breakdown of what to consider when designing Uber from an LLD perspective: 𝐑𝐢𝐝𝐞 𝐌𝐚𝐭𝐜𝐡𝐢𝐧𝐠 𝐀𝐥𝐠𝐨𝐫𝐢𝐭𝐡𝐦: 𝐊𝐞𝐲 𝐂𝐨𝐦𝐩𝐨𝐧𝐞𝐧𝐭𝐬: Efficient matching of riders with drivers. 𝐃𝐚𝐭𝐚 𝐒𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐬: Utilize priority queues or min-heaps to manage available drivers based on distance, ETA, or surge pricing. 𝐂𝐨𝐧𝐜𝐮𝐫𝐫𝐞𝐧𝐜𝐲: Implement locking mechanisms to handle concurrent ride requests effectively. 𝐋𝐨𝐜𝐚𝐭𝐢𝐨𝐧 𝐓𝐫𝐚𝐜𝐤𝐢𝐧𝐠: 𝐆𝐏𝐒 𝐈𝐧𝐭𝐞𝐠𝐫𝐚𝐭𝐢𝐨𝐧: Design a service to continuously update driver and rider locations. 𝐃𝐚𝐭𝐚 𝐇𝐚𝐧𝐝𝐥𝐢𝐧𝐠: Ensure minimal latency and accurate location updates using real-time data streaming technologies like Kafka. 𝐓𝐫𝐢𝐩 𝐌𝐚𝐧𝐚𝐠𝐞𝐦𝐞𝐧𝐭: 𝐒𝐭𝐚𝐭𝐞 𝐌𝐚𝐧𝐚𝐠𝐞𝐦𝐞𝐧𝐭: Use a state machine to handle various trip states (e.g., requested, accepted, in-progress, completed). 𝐃𝐚𝐭𝐚𝐛𝐚𝐬𝐞 𝐃𝐞𝐬𝐢𝐠𝐧: Opt for a NoSQL database like Cassandra for high write throughput and eventual consistency. 𝐏𝐫𝐢𝐜𝐢𝐧𝐠 & 𝐏𝐚𝐲𝐦𝐞𝐧𝐭𝐬: 𝐃𝐲𝐧𝐚𝐦𝐢𝐜 𝐏𝐫𝐢𝐜𝐢𝐧𝐠 𝐄𝐧𝐠𝐢𝐧𝐞: Design a scalable engine to calculate ride fares dynamically based on demand, distance, and time. 𝐏𝐚𝐲𝐦𝐞𝐧𝐭 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠: Ensure secure and reliable payment gateways, incorporating retry mechanisms for failed transactions. 𝐒𝐜𝐚𝐥𝐚𝐛𝐢𝐥𝐢𝐭𝐲 & 𝐅𝐚𝐮𝐥𝐭 𝐓𝐨𝐥𝐞𝐫𝐚𝐧𝐜𝐞: 𝐌𝐢𝐜𝐫𝐨𝐬𝐞𝐫𝐯𝐢𝐜𝐞𝐬 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞: Break down the monolithic Uber system into modular services, each handling a specific responsibility (e.g., ride matching, payments). 𝐑𝐞𝐬𝐢𝐥𝐢𝐞𝐧𝐜𝐞:Implement circuit breakers and fallback strategies to handle failures gracefully. 𝐀𝐏𝐈𝐬 & 𝐈𝐧𝐭𝐞𝐫𝐟𝐚𝐜𝐞𝐬: 𝐑𝐄𝐒𝐓𝐟𝐮𝐥 𝐀𝐏𝐈 𝐃𝐞𝐬𝐢𝐠𝐧: Clearly define the APIs for driver and rider interactions, focusing on statelessness and scalability. 𝐈𝐧𝐭𝐞𝐫𝐟𝐚𝐜𝐞 𝐃𝐞𝐬𝐢𝐠𝐧: Ensure the design supports both synchronous and asynchronous communication between services. 𝐋𝐨𝐠𝐠𝐢𝐧𝐠 & 𝐌𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠: 𝐑𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐌𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠: Integrate tools like Prometheus and Grafana for monitoring system health and performance metrics. 𝐂𝐞𝐧𝐭𝐫𝐚𝐥𝐢𝐳𝐞𝐝 𝐋𝐨𝐠𝐠𝐢𝐧𝐠: Implement a centralized logging service to trace issues and analyze system behavior during failures.

4

Process vs Thread Processes and threads are both mechanisms provided by operating systems to allow concurrent execution of multiple tasks, but they have different characteristics: 1. Process: - A process is an independent execution environment that runs as a separate instance of a program. Each process has its own memory space, resources, and execution context. - Processes are isolated from each other, meaning that one process cannot directly access the memory or resources of another process without using inter-process communication (IPC) mechanisms. - Starting a new process typically involves more overhead compared to creating a new thread, as the operating system needs to allocate resources and set up a separate memory space for the process. - Processes are more robust in terms of fault isolation. If one process crashes, it typically does not affect other processes. 2. Thread: - A thread is the smallest unit of execution within a process. Multiple threads within the same process share the same memory space and resources. - Threads within the same process can communicate directly and share data without using IPC mechanisms. - Creating a new thread is generally faster than creating a new process, as threads share resources and memory within the same process. - Threads are lighter weight than processes, as they share resources and memory. However, this also means that if one thread crashes, it can potentially affect other threads within the same process. In summary, processes provide stronger isolation and fault tolerance but come with higher overhead, while threads offer lower overhead and efficient communication but are more vulnerable to crashes and errors. The choice between using processes and threads depends on the specific requirements of the application and the trade-offs between isolation, performance, and resource utilization. #ruby #rubyonrails

17

🔍 Deep Dive: RAFT and 2PC in YugabyteDB 🔍 Let's explore how RAFT consensus and Two-Phase Commit (2PC) work together in YugabyteDB's sharded architecture! 🧱🔀 RAFT Consensus Algorithm: 🔄 RAFT ensures replication within each shard: 1. Leader Election: 👑   - Each shard has a leader and followers   Example: Shard 1 with 3 replicas: A (leader), B, and C (followers) 2. Log Replication: 📝   - Writes go to the leader (A)   - A replicates to B and C   - Commit after majority acknowledgment 3. Safety: 🛡️   - If A fails, B or C can become the new leader   - New leader ensures up-to-date log before accepting writes Two-Phase Commit (2PC): 🔒 2PC manages transactions across multiple shards: 1. Prepare Phase: 🤔   - Transaction coordinator sends prepare message to shards   - Each shard leader checks commit possibility   - Positive responses are logged 2. Commit Phase: ✅   - All positive? Coordinator sends commit message   - Any negative? All shards abort Example: 💰➡️💰 Transferring money from Shard 1 to Shard 2: 1. Coordinator sends "prepare" to both shards 2. Shards check if they can process the transfer 3. If both agree, coordinator tells both to commit 4. Changes applied using RAFT within each shard The RAFT + 2PC combo ensures: - Strong consistency within shards 💪 - ACID properties across shards ⚗️ - High availability and fault tolerance 🏋️‍♀️ Join the Journey with Vikas K. 🔥 Follow for a World of Insightful Content! #YugabyteDB #DistributedSystems #RAFT #2PC #DatabaseTechnology #DataEngineering #Consistency #FaultTolerance

5