Antoni Rosinol’s Post

#Devin, the world's first AI software engineer ---------------------------------------------------- 🕹US-based startup #Cognition has unveiled #Devin, an artificial intelligence (AI) software engineer that promises to revolutionize the way code is written, debugged, and deployed. 🕹 Devin is a software development assistant that surpasses mere code completion and suggestions. Unlike large language models (LLMs) such as #OpenAI's #ChatGPT or Google's #Gemini, Devin can autonomously handle the entire process of building and launching full-fledged software applications. 🕹 The Founders Fund-backed startup claims that the AI assistant is the new state-of-the-art on the SWE-bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on #Upwork. Devin can plan and execute complex engineering projects, 🕹 involving thousands of decisions. It learns from its experiences, recalls relevant context, and even fixes its own mistakes. #Devin #llm #artificialintelligence #neuralnetworks #lora #qlora #development

We're #hiring at Sobek AI! Come join us as a Founding Full Stack Engineer. Has been a wonderful journey so far, and we're just getting started. Details here: https://sobek.ai/career/ #hiring #artificialintelligence #generativeai #biopharma #startup #genai #techstartup

The First AI Engineer 𝐃𝐄𝐕𝐈𝐍 is here! 𝑻𝒉𝒆 𝑺𝒐𝒇𝒕𝒘𝒂𝒓𝒆 𝑻𝒆𝒂𝒎 𝒐𝒇 𝒕𝒉𝒆 𝑭𝒖𝒕𝒖𝒓𝒆 𝒊𝒔 𝒉𝒆𝒓𝒆! Cognition AI, a recently formed AI startup announced a fully autonomous AI software engineer called “Devin”. I have seen 😳 the demos and mind absolutely blown! 🤯 ⏳ WHAT WAS EXPECTED IN 3 OR MORE YEARS HAS BEEN ACHIEVED IN 6 MONTHS. 💸 𝘾𝙤𝙜𝙣𝙞𝙩𝙞𝙤𝙣 𝘼𝙄 is backed by Peter Thiel’s Founders Fund and other tech industry leaders, including former Twitter executive Elad Gil and Doordash co-founder Tony Xu. 💵 Devin has passed interviews and completed projects on Upwork, and then some for developers at Cognition AI: 1. Devin works autonomously in its own editor, workspace and browser and can self-debug with "print" statements and move projects forward without any assistance from outside. 2. Devin can install packages and fault-find and correct when open source packages don't work and re-install as required. 3. Devin can write test cases to debug and fix complex bugs and refactor code. 4. Devin can train other AI's without any assistance and can report back status / progress. 5. Devin can take instruction - specification - and deliver what is expected (e.g. blog posts, this is the easiest thing I saw it do). I want my own Devin! 𝙞𝙢𝙖𝙜𝙚: 𝘢𝘶𝘵𝘩𝘰𝘳 𝘤𝘳𝘦𝘢𝘵𝘦𝘥 𝘢𝘯𝘥 𝘱𝘰𝘴𝘵𝘦𝘥 𝘩𝘦𝘳𝘦 𝘦𝘢𝘳𝘭𝘪𝘦𝘳 𝘪𝘯 2024. #AIEngineer #AGIFirstContact #AGI #SoftwareEngineering

tl;dr: Founding MLE/MLOps, $200k–380k/year + equity, relocation to London While hopes and fears are brewing around Ilya Sutskever’s next steps, we got you a safety supervising intelligence job opening. Atla is a London-based startup that is building AI for automatic evaluation and testing of other AI products. More specifically, they are training LLMs to assess other models while also providing a handy API for genAI app developers. Why is this necessary? Well, for example, to make sure that your model won’t suggest its users to keep cheese on their pizza with superglue 🤨 Why LLMs? Because assessing models with human efforts is very expensive and hard-to-scale. General-purpose models aren’t doing too well on this front, and Atla is one of the pioneers in developing supervisor models. They received $5M in seed funding — this was the biggest round in the S23 batch — from Y Combinator, Creandum and founders of Reddit and Cruise. We are looking for a Founding ML Engineer to join their team. You will be, erm, supervising the development of fine-tuning frameworks and LLMs, collaborating with the research team on data generation, and doing other stuff that important ML experts do 🧑💻 Let’s say, you: ✨ have managed multi-instance clusters for parallel training across hundreds of GPUs/TPUs — preferably using PyTorch and Kubernetes; ✨ have launched, served and optimized large ML models; ✨ are very cool! For example, you’re from a top uni and have publications, or you have worked in Stability or Meta, or maybe your Kaggle rating is through the roof. We want to know! ❤ Since they’re in an early stage, the equity share is significant. Your visa will be sponsored and relocation arranged through the UK AI Futures programme. The annual salary range is $200k-380k. Direct all your CVs, questions and referrals to Nikita 📝

Since we do a lot of hiring at AxcelerateAI, we often get asked, "What does it take to become a good AI engineer?" Here are some key areas we focus on when evaluating candidates: 1️⃣ Mathematics is Non-Negotiable: A strong grasp of calculus, probability, and statistics is essential. Without a solid foundation in these areas, it’s nearly impossible to truly understand how machine learning algorithms work. And if you don't understand the algorithms, you can’t effectively train, fine-tune, or optimize them. Math is the language of AI, and it’s crucial for building models that actually perform. 2️⃣ Strong Programming Skills: Knowing Python is a start, but it's not enough. You also need to master best coding practices, data structures, and algorithmic analysis. Writing efficient code is crucial when working with large datasets and training complex models. The ability to optimise code for performance and scalability is what separates a good engineer from a great one. 3️⃣ Thinking Mathematically: Successful AI engineers don’t just rely on intuition—they think in terms of equations and logic. Translating mathematical concepts into structured problem-solving steps is key to building AI systems that work reliably. Mathematical thinking allows you to break down complex problems into manageable pieces and approach them with clarity. We’ve also developed an internal, more detailed roadmap for becoming an AI engineer. If you’re interested in learning more, feel free to DM me! #MachineLearning #AxcelerateAI #CareerGrowth

Backend Engineering IS NOT AI Engineering. 🤐 “𝑾𝒉𝒂𝒕 𝒊𝒔 𝒕𝒉𝒆 𝒇𝒊𝒏𝒆 𝒍𝒊𝒏𝒆 𝒃𝒆𝒕𝒘𝒆𝒆𝒏 𝒃𝒂𝒄𝒌𝒆𝒏𝒅 𝒆𝒏𝒈𝒊𝒏𝒆𝒆𝒓𝒊𝒏𝒈 𝒂𝒏𝒅 𝑨𝑰 𝒆𝒏𝒈𝒊𝒏𝒆𝒆𝒓𝒊𝒏𝒈?” I was asked this question yesterday at an interview, and my mind froze for a couple of minutes. I realized that working in multiple startups has made me wear many hats 🤠 – from software and backend engineering to AI expertise and even the CTO role. This journey has taught me the art of adaptability, always ready to jump into any challenge that comes my way. But that doesn’t mean that’s the only path of evolution you can have as an engineer. There are also drawbacks to following a multi-perspective technical journey: ♦️ You get into too many roles and responsibilities and forget the details; ♦️ You get distracted by soaking up a lot of information in different domains; ♦️ Sacrifice quality to accumulate quantity. Finally, my answer to this interview was a personal experience opinion. There’s a fine line between backend and AI engineering, and it depends on the following question: 𝐖𝐡𝐚𝐭 𝐚𝐫𝐞 𝐰𝐞 𝐛𝐮𝐢𝐥𝐝𝐢𝐧𝐠❓ 𝐀𝐧 𝐌𝐕𝐏 𝐨𝐫 𝐚 𝐏𝐫𝐨𝐝𝐮𝐜𝐭 ❓ In an MVP setting, blending backend and AI engineering is more common. Quick, functional prototypes are key, requiring a broad skill set. However, for an advanced product, the distinction is clearer. ⚙️ Backend engineering focuses on building a robust, scalable infrastructure - the foundation essential for AI implementation. 🤖 AI Engineering, meanwhile, involves developing intelligent algorithms that leverage this infrastructure, transforming data into actionable insights and advanced features. There's indeed an overlap in technical skills, but backend and AI engineering divide in their core objectives and contributions to the product’s lifecycle. The backend lays the groundwork, ⛓️ while AI brings the innovation. 🧠 𝐋𝐞𝐭 𝐦𝐞 𝐤𝐧𝐨𝐰 𝐲𝐨𝐮𝐫 𝐭𝐡𝐨𝐮𝐠𝐡𝐭𝐬! What do you consider to be a role differentiator between a backend engineer and an AI engineer? #aiengineering #artificialintelligence #aiengineer

🚀 350 Promising Startups Hiring Right Now! 🚀 I've been reading up on hundreds of startups and have compiled a shortlist of those showing great potential and currently hiring. If you're looking for new ways to find hidden job opportunities, keep an eye on the comments section. I'll be sharing ideas periodically that you can try. StyleAI (all-in one marketing platform) Obviously AI (no-code AI insights) HockeyStack (no-code analytics) Anarchy Labs (Gitlab for LLM teams) BigSpring AI (AI-powered revenue acceleration) Shipday (AI delivery for businesses) Cleanlab (data-centric AI software) Opus Clip (AI video clipping & editing tool) PermitFlow (permit app & mgmt software) Play (conversational voice AI) BoomersHub (senior care referral service provider) Slope (B2B payments platform) Sybill (AI body language reader) Pylon (AI customer operations platform) Credo AI (AI governance platform) Workyard (workforce mgmt solution) Gradient (contextual AI for workload automation) Numenta (deploy large AI models on CPUs) Kira Learning (AI solutions for teachers & schools) Reflex (open-source full-stack Python framework) Foxglove (robotics development platform) Lakera (safeguards AI apps) 🔭 Galileo (algorithm-powered LLMOps platform) Dynamo AI (compliant-ready AI for enterprise) Coram AI (AI-driven software for security cameras) Hivemapper (map data products) Resemble AI (Generative Voice AI toolkit) Geomagical Labs (AI-powered digital experience) Acryl Data (streaming metadata platform) Squint (AI data entry) Stronghold (virtual payment networks via API) Socket (developer first security platform) Improvado (automated revenue data platform) Anrok (tax compliance infrastructure) LangChain (orchestration layer for AI apps) WorkOS (enterprise ready developer platform) Freshpaint (high-performance marketing compliance) Dusty Robotics (construction layout robots) Clumio (data protection software) Lightning AI (AI development platform) See comments for even more promising startups that are hiring. #Startups #JobSearch #Networking #CareerOpportunities #Innovation

When I first read this tweet, I didn't agree with it. I believed that programming skills weren't crucial enough to be a determining factor in a startup's failure. However, my perspective changed drastically when I embarked on my own startup journey last year. I learned firsthand that the initial year of a startup is pivotal, laden with immense pressure, high expectations, and financial crises, not to mention instability. Hiring technical staff at the outset can create significant financial strain, potentially leading to a startup's downfall. In the early stages of my startup, our objective was to identify potential asteroids using machine learning models, necessitating the expertise of a machine learning engineer and a data scientist. Finding such talent, proficient in ML models and knowledgeable about space data, was a rare and costly endeavor. So, I took it upon myself to learn machine learning. With a background in statistics and mathematics, where I had studied Probability Theory, Descriptive Statistics, Bayesian Statistics, Linear Algebra, Calculus, Discrete Mathematics, and algorithm complexities during my masters, including programming and software engineering, I found these concepts immensely helpful in grasping ML models. Despite not having used these skills in my banking career, they proved invaluable in my startup journey. I chose not to wait for funding to hire an ML engineer; instead, I embarked on a self-learning journey. This decision not only equipped me with vital knowledge and skills but also provided stability to my startup, allowing us to make significant progress in our first year. It was a challenging period, involving 80-hour workweeks with no holidays or weekends off, but it was worth it. The advent of AI and its applications has been a game-changer, significantly reducing the cost and time of startup operations. Even for startups that are not AI-centric, leveraging AI tools allows for rapid development. Post the AI boom, I revisited the tweet and realized that, thanks to numerous AI applications and the availability of low-code and no-code tools, programming has become much more accessible, even with minimal programming knowledge. One thing I'm certain of is that if someone is starting a tech startup in the next five years, they must learn AI/ML skills. And by ML skills, I don't just mean programming skills but also a deep understanding of mathematics and statistics concepts. #artificialintelligence #machinelearning #programing #startup #spacetech #spaceexploration

𝗗𝗲𝘃𝗶𝗻: 𝗥𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝗶𝘇𝗶𝗻𝗴 𝗔𝗜 𝗼𝗿 𝗝𝘂𝘀𝘁 𝗔𝗻𝗼𝘁𝗵𝗲𝗿 𝗧𝗼𝗼𝗹? 💻 Devin, touted as a potential replacement for software engineers and programmers, has stirred quite a buzz in the tech community. But is it truly a game-changer, or just another iteration of existing AI tools? 👉🏽 I believe Devin is essentially a more refined version of AutoGPT, boasting a sleek UI and user-friendly interface. However, its effectiveness hinges heavily on one critical factor: context. ▶️ Context is king in the world of AI. The more context Devin receives, the sharper its responses become. Yet, capturing the intricacies of real-world scenarios—such as those encountered in app development—poses a significant challenge. 🌏 Software engineering isn't just about writing code; it's about navigating a labyrinth of meetings, discussions, and evolving requirements. Each day brings new challenges that demand nuanced solutions—a feat not easily achieved by AI alone. ✨ While AI continues to evolve, the race for increased token counts reflects the industry's pursuit of providing AI models with richer context. Yet, even with advancements like Gemini Pro 1.5 boasting 10M tokens, Devin's role remains more akin to that of a coding assistant rather than a replacement for human engineers. 🔥 Cognition Labs' marketing strategy capitalizing on Devin's purported capabilities as a software engineer replacement has undeniably captured attention. However, amidst the hype, it's crucial to recognize Devin's limitations and its evolving role in the developer ecosystem. 👏🏽 Despite the skepticism, Devin has succeeded in making a name for itself in the crowded landscape of AI tools. Even Reddit, Inc. has dedicated extensive discussions to dissecting Devin's potential impact. ✅ Yet, amidst the buzz, one can't help but ponder: if Devin is meant to replace human software engineers, why is Cognition Labs hiring human engineers for Devin's development? Food for thought indeed. ❓ So, should we brace ourselves for an AI dominated future in software engineering? Perhaps not just yet. While Devin may offer valuable assistance, the human touch remains irreplaceable at least for the foreseeable future. ______ #ai #devinaI #softwareengineering #aitools #contextualaI #geminipro #codingassistant #techindustry #aIlimitations #cognitionlabs #futureoftech #artificialintelligence #aisoftwareengineer