Amanj Sherwany

One of the skills that you need to invest in as a data scientist is learning Docker. Docker is a containerization platform that can be used for a variety of purposes, including machine learning and data science. We have published a series of articles on To Data & Beyond to start in Docker for data science: ➡ Docker for Data Science Projects: A Beginner-Friendly Introduction Link: https://lnkd.in/dQ-D33J6 ➡ 11 Docker Container Images for Generative AI & ML Projects Link: https://lnkd.in/deXF_CmE ➡ Building your first Docker container for Machine Learning Applications Link: https://lnkd.in/dsYQdvYG ➡ Industry-Wide Best Practices For Using Docker for Data Science Link: https://lnkd.in/dd3vNGX9 ➡ 10 Docker Commands for 90% of Containerization Tasks Link: https://lnkd.in/dCSQsvSC ----------------------------------------------------------- ➡ Do not miss the 50% discount on To Data & Beyond yearly subscription: https://lnkd.in/dZ2Sjkhk ➡ A Beginner to Upper Intermediate Data Science Roadmap with detailed resources https://lnkd.in/deEBAAgw

1,591

21 Comments

More engineers ≠ more progress. Sometimes, it’s the opposite. Throwing people at a project doesn’t speed things up. It adds complexity. - More meetings - More handoffs - More syncing - More coordination overhead And every new hire slows someone else down. Because onboarding pulls seniors off real work. Before you hire, fix the system. Tighter feedback loops, better docs, and clear ownership often unlock more speed than headcount ever will. Fix the system first; more people won’t save a broken setup.

2,128

163 Comments

I spent the first few months in 2023 job hunting in the AI/Data Science market. I had to go through the process unexpectedly again at the beginning of 2025. I had collated my interview questions in 2023 and I found this a very useful starting point while preparing for the second time, especially since unlike SDE interviews, ML interviews can be completely different at different companies, and with so much to review (Resume projects, leetcode, classical ML, deep learning, CV/NLP, pytorch/pandas/numpy coding ...) it can quickly become daunting. So I thought I would share my (updated) document for anyone who's going through the interview grind or planning to start it in the near future. I hope it helps you in landing your next role. https://lnkd.in/g4urc5CV Some other things I observed during my job-hunt: - My IIT Bombay degree helped me immensely in getting interviews, so USE your network - Having a CVPR research paper impressed my interviewers - Using GenAI coding tools has negatively affected my coding skills, particularly with python libraries like pandas, numpy, PyTorch, so it's probably time to take a break from the 'tabbing' habit. #jobhunt #datascience #machinelearning #deeplearning

196

9 Comments

Most of us use agentic coding tools to build products. But what happens when you use them to build the operating system for your life? At this month’s AI Tinkerers: Kitchener–Waterloo, Artem Zhutov will demo how he built a personal AI assistant using Cursor and Claude Code—not for codegen, but for context. This assistant tracks goals, manages routines, and automates context engineering so you can stay in flow across projects. He’ll walk through live examples of: ⚙️ Automatic context engineering – how your AI can remember what happened three days ago and surface it when it matters 🎯 Goal & experiment tracking – integrating your personal knowledge base with agentic workflows 🌅 Morning / evening routines – structured reflection loops grounded in Atomic Habits 🧩 Modular agentic workflows – reusable building blocks that turn intent into execution If you’re already shipping with AI tools, this talk shows how to turn those same workflows inward—using agentic systems to handle your meta-work: thinking, prioritizing, and context-keeping. For solopreneurs, founders, and students, it’s a framework for building personal infrastructure for focus—without relying on heavyweight productivity stacks. As always, this is a builders-only room: no pitches, no fluff, just real demos and real conversations. 📍 Builders Club – 165 King St. W, Suite 301 ⏰ Monday, October 20th at 6pm 🎟️ Apply to join the builder roster: link in the comments Jesse Rodgers Mike Bird Kris Braun

24

3 Comments

Large Language Models are getting smarter AND smaller. This video breaks down how AI researchers have made this possible. 00:00 Overview 00:20 Scaling Law 1.0 01:48 Using smaller weights 03:44 Increasing model training 06:05 Google Titans 07:40 What's after Transformers? 08:45 Neuromorphic Computing #LLMs #AI #Research

317

10 Comments

Docker just launched Model Runner and it’s going to change how we run LLMs. 📦 Think of it as: 🧠 LLM inference + 🔧 Docker CLI + 💨 GPU speed, all local. 🎯 Top Highlights: - Seamless integration with your Docker workflow - Built-in support for llama.cpp & OpenAI-compatible APIs - GPU acceleration on Apple Silicon ⚡ - Model packaging using OCI Artifacts (finally, a standard!) 📦 - Pull-and-run from Docker Hub. Push your own models soon. 🌍 Use It For: - Private LLM experimentation - Speeding up app iteration - Avoiding cloud lock-in - CI pipelines with AI inference baked in 🤝 Supported by leaders in the AI space: Hugging Face, Google, Qualcomm, Continue, Spring AI #ModelRunner #GenAIDev #EdgeAI #DeveloperTools #AIPrivacy

99

4 Comments

Mass communication research in Pakistan: what’s missing and why it matters By Prof. Dr. Taimoor ul Hassan Mass communication shapes public opinion, culture, and governance—yet research in Pakistan remains fragmented and underdeveloped. Despite the growth of media studies departments, our academic work often falls short in quality, relevance, and innovation (Yousaf & Ali, 2020). A major challenge is overreliance on Western theories that don’t fully capture Pakistan’s unique social, cultural, and political dynamics. There is an urgent need for locally grounded frameworks that reflect our realities (Raza & Khan, 2015). Methodological weaknesses persist. Many studies rely on small, unrepresentative samples and basic analyses. Advanced qualitative methods and mixed approaches are rarely employed (Aslam, 2016). Coupled with limited access to resources and research funding, this results in repetitive studies that lack impact (Jamil, 2021). Another critical gap is the disconnect between academia and the media industry. Research findings rarely influence newsroom practices or policy decisions, and practitioners have limited engagement with scholarly work (Raza & Khan, 2015). Language barriers further restrict the reach of research. Most publications are in English and inaccessible to journalists and media professionals who work primarily in Urdu or regional languages (Jamil, 2021). Finally, emerging areas like digital media, AI, misinformation, and media economics remain largely unexplored, while voices from rural and marginalized communities are underrepresented (Yousaf & Ali, 2020). To transform mass communication research in Pakistan, we need curriculum reform, stronger research culture, greater industry collaboration, and accessible multilingual publishing. Only then can we produce scholarship that truly informs media practices and public discourse. Copywriting: Dr Taimoor ul Hassan , UCP. --- #MassCommunication #MediaResearch #Pakistan #AcademicResearch #MediaIndustry #DigitalMedia #MediaStudies #ResearchCulture #Journalism #CommunicationStudies #MediaDevelopment #AIinMedia #MediaInnovation #InclusiveResearch ---

12

2 Comments

💡 𝗔𝗜 𝘄𝗿𝗶𝘁𝗲𝘀 𝘁𝗵𝗲 𝗰𝗼𝗱𝗲. 𝗜 𝗷𝘂𝘀𝘁 𝘃𝗶𝗯𝗲. I wrote a full piece about this topic, including: ✅ Why AI is raising the bar for everyone ✅ Why juniors will struggle more to break in ✅ What skills truly future-proof your career ✅ And why “thinking like an engineer” matters more than ever 🧠 Read the full post here → https://lnkd.in/dwzCHK-u Let me know what you think.

16

3 Comments

You gave one interview. You failed. You stopped preparing. Because you were chasing 100% precision. 4.5 years ago, I bombed "most recent project" interview round of a top 10 company. Not just bombed, I imploded. I couldn’t explain my own project clearly. I fumbled through basic design questions. I froze when asked about trade-offs and requirements. Like: - availability - single point of failure - why did I choose haproxy over nginx - how did I handle failure and build fault tolerance Because I never thought about it in this "detailed manner" while building it (and most of us build this way) It wasn’t only knowledge problem, It was a self-awareness problem too. Instead of hiding from the failure, I did something which changed the outcome for my next interviews: - I spent 8–10 hours deciphering my own project, a system I had already built and shipped to production. - I wrote down, physically, not mentally, every shortcoming I exposed in that interview. - I created a 2-week plan to attack each weakness one by one. That brutal self-reflection turned my worst interview into the foundation for offers later on. But most people don’t get there. They fail once. They stop trying. They waste months or years thinking they "aren't good enough." ML algorithms don't give 100% precision. Neither will your interviews. Neither will your life. If such algos, which powers self-driving cars, cancer diagnostics, and trillion-dollar tech, can't achieve perfection... Why are YOU demanding 100% precision before allowing yourself to keep going? Rejection is not the signal that you're "done." Rejection is the raw material for your next level. If you just failed an interview: Step 1: Reflect brutally. What specifically went wrong? Write it down, no excuses. Step 2: Request feedback. If the recruiter shares insights, amazing. If not, you already know where you slipped. Step 3: Create a 2–3 week sprint to improve the broken parts, not everything at once. Step 4: Normalize imperfection. Precision is a myth. Progress is real. Failing once doesn’t make you unworthy. Stopping after failing does. Remember these 5 points (which changed the way I look at interviews now): 1/ Failure is not final : it's just feedback. If you treat rejection like a death sentence, you’re letting fear run your career. 2/ Mental lists are a lie : if you don't physically write down your gaps, you won't fix them. Your brain will quietly ignore what feels uncomfortable. 3/ Timing is everything : the 2–3 weeks after a rejection are golden. That's when the pain is fresh, the lessons are vivid, and your hunger is highest. USE IT. 4/ No one is keeping score : you're the only one obsessing over that one failed interview. Companies move on fast. So should you. 5/ Clarity compounds : when you can explain your past work simply and confidently, you not only ace interviews, you become a better engineer and leader. Feeling stuck, and need help? DM me COACH.

69

18 Comments

I’ve been fortunate to interview with a range of companies—from FAANG giants to early-stage NYC startups that are building fast and dreaming big. The interview pattern in both is entirely different—and both bring out different strengths in you. 🔍 𝗙𝗔𝗔𝗡𝗚 𝗜𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄 𝗣𝗮𝘁𝘁𝗲𝗿𝗻(SWE) : Online Assessment – DSA and coding questions under timed constraints. Technical Rounds – Multiple 1:1 rounds with deep focus on data structures, algorithms, and problem solving. System Design – For experienced roles, a round dedicated to designing scalable, distributed systems. Behavioral – Leadership principle-based questions, focused on collaboration, ownership, and impact. 𝗘𝘃𝗲𝗿𝘆 𝘀𝘁𝗲𝗽 𝗶𝘀 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱, 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘇𝗲𝗱, 𝗮𝗻𝗱 𝗳𝗼𝗰𝘂𝘀𝗲𝗱 𝗼𝗻 𝗵𝗼𝘄 𝘄𝗲𝗹𝗹 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸 𝗮𝘁 𝘀𝗰𝗮𝗹𝗲. ⚡ 𝗦𝘁𝗮𝗿𝘁𝘂𝗽 𝗜𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄 𝗣𝗮𝘁𝘁𝗲𝗿𝗻(SWE): Intro Call – Often with the founder or CTO, informal yet meaningful. Technical Round – Focused more on building real features, reading codebases, or solving practical problems. Live Coding / Take-home Task – Build a small app, debug a service, or improve performance. Culture Fit / Vision Discussion – How you handle ambiguity, take initiative, and thrive in a fast-moving, multi-role environment. 𝗠𝗼𝗿𝗲 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝗮𝗹, 𝘄𝗶𝘁𝗵 𝗮 𝗳𝗼𝗰𝘂𝘀 𝗼𝗻 𝗵𝗼𝘄 𝘆𝗼𝘂 𝗯𝘂𝗶𝗹𝗱 𝗿𝗮𝘁𝗵𝗲𝗿 𝘁𝗵𝗮𝗻 𝗷𝘂𝘀𝘁 𝘄𝗵𝗮𝘁 𝘆𝗼𝘂 𝗸𝗻𝗼𝘄. 🌟 My take? FAANG interviews help you master the fundamentals and think long-term. Startup interviews prepare you to solve problems today. Both are valuable. Both help you grow. Curious to hear from others—what patterns have you seen, and what’s helped you prepare best? #FAANG #StartupLife #InterviewExperience #SoftwareEngineering #TechInterviews #CareerGrowth #MyTake #NYC

40

1 Comment

PART 2: HOW WE IMPLEMENTED THE ASSET OPTIMIZATION PACKAGE THAT SAVED OUR APP 67% 🔥 Since my post went viral yesterday (THANK YOU!), my inbox is FLOODED with "HOW DID YOU DO IT?!" messages! 📱💪 THE RESULTS WERE INSTANT: 143 unused PNGs/SVGs automatically removed! 🧹 76 duplicate assets we didn't even know existed! 🤦‍♂️ All remaining images compressed at build time 📉 Generated type-safe references (goodbye String typos!) 👋 BONUS TIP: Integrate this into your CI/CD pipeline and watch your build size shrink with EVERY DEPLOYMENT! 🚀🚀 The performance gains were most dramatic on entry-level Android devices in emerging markets – exactly where we needed to improve! 📊 Has anyone else tried this yet? Drop your results below! 👇 #FlutterDev #MobileOptimization #PubDev

183

18 Comments