Shaw Talebi's Linkedin Analytics

Although we often see the highlight reel of entrepreneurship on social media, the reality couldn’t be more different... It took me 15 months of failure and volatility to (finally) match my corporate salary ($10k/mo). Here are the 4 hardest lessons I learned on my journey: 1) It’s not real until the money is in the bank. In the beginning, it was easy to get excited about "promising signals" or "potential clients" just for it to (usually) fall flat. This emotional rollercoaster can be draining, so it's best to defer your excitement until after the money hits the bank (if getting excited at all) 2) Employee → Entrepreneur mindset shift Don’t ask, “What work can I do?” or “What services can I provide?”. Instead, ask, “What problems can I solve?” The latter is the key driver of value, the former is just a means to an end. 3) Focus more. Every time I got more specific on the problem I'm solving and who I'm solving it for, my life got easier (and I ended up making more money). Less is more. 4) “Real artists steal” It took me way too long to create my AI bootcamp, because I didn't want to be "yet another" YouTube with a course. But the reason this is so popular is because it works. Don’t let trying to be original and clever keep you poor. See what’s working for others and adapt it to your situation. In a recent YouTube video, I share behind-the-scenes of my journey to this milestone and key lessons learned along the way. 👉 Check it out here: https://lnkd.in/gMwyqfeR

78 AI project ideas... (built by students in my AI Builders Bootcamp) 👇 1. AC Milan Reminder 2. AI Accent Coach 3. AI Executive Coach 4. AI Text Detector 5. Agentic Doc Analyzer 6. Applicant Matching System 7. Arxiv Extractor & Summarizer 8. Arxiv Paper Retriever 9. Automated AI Newsletter 10. Automated Birthday Emailer 11. Automated Email Reminders 12. BGP Anomaly Detection 13. BGP Data Stream 14. Bluesky QA Bot 15. Bulk File Copier 16. Catering Menu Creator 17. Chess Grandmaster Ranking 18. Churn/Fake News Classifiers 19. Coaching & Therapy Assistant 20. Competitor RFP Scraper 21. Cover Letter Enhancer 22. Credit Risk Modeling 23. Customer Review Analysis 24. Document Parser 25. Ebay iPhone Scraper 26. Edge Network Monitoring Assistant 27. Email Inbox Organizer 28. Excel Data Pipeline 29. Fantasy Football ETL 30. File Organizer 31. Football SuperLeague Leaderboard 32. Foundational RAG System 33. Fraud Detection 34. Geopolitical Market Analyzer 35. Job Board Scraper 36. John Lennon Chatbot 37. LI Internship Scraper and Emailer 38. LLM-powered Anomaly Detection 39. Meeting Emotion Analysis 40. Mindbody Scraper 41. News Feed Aggregator 42. No-show Email Automation 43. Outlook Data Pipeline 44. PDF Summarizer 45. Patent Extract/Scraper 46. Personalized Mortgage Rate Emailer 47. Predicting Housing Prices 48. Predicting Natural Gas Prices 49. Product Data ETL 50. Python Project Summarizer 51. Radio Station Watcher 52. Real Estate Image Finder 53. Recipe Recommender System 54. Reddit Post Classifier 55. Report Image Extractor 56. Research Paper QA Bot 57. Resume Matcher 58. Resume Summarizer 59. Root Cause Analysis Tool 60. Router Config Generator 61. STNet Paper QA 62. Sales ETL & Prediction 63. Satirical Headline Generator 64. Stock Analysis Tool 65. Stock Picker 66. Stock Price Alert System 67. Stock Price Prediction 68. Stock Report Generator 69. Stock Sentiment Analysis 70. Student Assessment Tool 71. Support Ticket Classifier 72. Textbook Chapter Splitter 73. Transcript Cleaner 74. Travel Itinerary Planner 75. YT Video Summarizer 76. iPhone Price Predictor 77. iTunes Library Analysis 78. MySQL Query Assistant -- If you've been procrastinating your awesome AI project ideas, join the upcoming cohort for direct guidance from me and a supportive community of peers :) 👉 Learn more: https://lnkd.in/gPHZTdWs

Although LLMs seem to perform any task we throw at them... ... this flexibility turns into unpredictability when engineering new AI apps. This is where evaluations (i.e evals) can help. In a recent YouTube video, I discussed 3 ways of evaluating LLM systems to ensure they (actually) do what you want. Type 1: Code-based = rule-based assessment of a model's output Type 2: Human-based = using human judgment (or behavior) to evaluate output quality Type 3: LLM-based = using LLMs to assess output quality 👉 Check out the full video here: https://lnkd.in/gZnF5awZ

LLMs have lowered the barrier to entry for building AI products. What used to take months of data curation and model training can now be done with an API call to your favorite LLM provider. Although going from 0 to 1 is easier than ever, going from prototype to production application is still not trivial. This Friday (Mar 21) at 10AM CST, ML engineer Jeremy ARANCIO will walk through a real-world LLM use case he built for one of his clients. Join us live and ask Jeremy your LLM development questions! 👉 Register here: https://lu.ma/84d5wgcu

The problem with “AI agents” is that no one seems to agree on what it means. Basically, anything involving automation or LLMs is at risk of being an AI agent. This is partly because the term has evolved over the past 2 years. Here are 3 distinct phases of “AI agents” I’ve noticed: Phase 1: LLM + Tools These were the first types of systems I heard of called AI agents. Agency comes in because we can give LLMs tools like API calls and search for interacting with the world in a self-directed way. Phase 2: LLM Workflows This is the most common definition I see these days. The idea is to define workflows that involve LLMs as nodes and decision-makers to perform complex tasks. Phase 3: End-to-end RL Training The latest iteration of agents are systems like OpenAI’s Operator, which are trained on end-to-end reinforcement learning. They can handle open-ended tasks and interact with real-world systems without much supervision. What are some other types of AI agents you’ve seen? -- If you liked this post, repost it and follow me (Shaw Talebi) for more

2025 is said to be “the year of AI agents” However, no one seems to agree on a single definition for these systems. Here’s my explanation of AI agents (without the hype)… Since people can’t agree on a single definition of an AI agent, most practitioners talk about agentic systems instead. In other words, thinking about agents on a spectrum of agency, ranging from no agency (e.g. rule-based system) to human-level agency. Here are some examples of that. Level 1: LLM + Tools 🤖🪛 Giving LLMs access to tools like a code interpreter, web search, or retrieval system allows them to gather information from the real world and perform actions in it. Level 2: LLM Workflows 🤖 ➡️ 🤖🪛 While giving an LLM tools makes it practical, it is still limited because it is forced to perform tasks in a one-shot fashion. We can overcome this by orchestrating several software components (code or LLM) to accomplish complex tasks. Level 3: LLM in a Loop 🤖🔄 Some tasks may not be predictable enough to solve with a pre-defined workflow (even with LLMs). This is where looping LLM calls can be powerful. This basic flow goes like this. Generate response from LLM → Evaluate response → if task is complete stop; else add feedback to context and call LLM Are you building any agents in 2025? -- 👉 If you liked this post, repost it and follow me (Shaw Talebi) for more

How I grew from 0 to 70,000 followers (on Medium and YouTube). There was never a master plan with my content creation. I did it mainly because 1) explaining a topic solidified my understanding of it and 2) I was a physicist who struggled with communicating to a non-technical audience. However, over the past four years, content creation has created opportunities that I never imagined, including freelance work, speaking gigs, and, most importantly, the ability to work for myself and do exactly what I want to do every day. Last week, I shared my secrets and hard-learned lessons in a live talk hosted by The Data Entrepreneurs ICYMI–the recording is now available on YouTube 👇 🔗 Check it out here: https://lnkd.in/gf8XJv8E

AI Agents present a new way of thinking about software... Rather than explicitly defining rules and business logic, agents involve giving LLMs the tools they need to solve problems. For example, instead of developing a rule-based customer support bot, you might give an LLM... 1. Context about your company 2. A search tool for support docs 3. The ability to escalate issues With today's powerful LLMs fine-tuned for function calling, such a naive system can get you much of the way there out of the box. In a new blog post, I explain how LLM tool-use works under the hood and build a simple YouTube video agent using OpenAI's new Agents SDK. 👉 Check it out: https://lnkd.in/gr_cttku -- Join 1,000+ AI builders and get 30 free project ideas: https://lnkd.in/gf7z2eab

10 (non-technical) skills I needed to learn to survive as a data entrepreneur. For tech people, going from the top 10% to the top 5% of coders is LESS valuable than going from the bottom 50% to the top 50% in key non-tech skills. Here are several that have moved the needle for me... 1) Explaining technical ideas to non-technical people 2) Talking to customers 3) Formulating business problems 4) Asking good questions 5) Validating product/service ideas 6) Responding to emails quickly 7) Sales 8) Warm outreach 9) Making a landing page 10) Copy writing What would you add here? -- 👉 If you liked this post, repost it and follow me (Shaw Talebi) for more

Shoutout to AI Builders Alumna Sangeeta Bahri for shipping her first AI app! Most retail investors don’t have access to the deep analysis and tools big firms use, leaving fewer quality investment opportunities for the rest of us. This is where “Equity Research” comes in… By combining real-world expertise and AI-driven analysis, Sangeeta helps everyday investors get expert-level analysis to make better investment decisions. She started by building various features of her product as weekly projects in the AI Builders Bootcamp. Today, she’s talking to customers and bringing a real solution to market. This is exactly why I started the boot camp — to help entrepreneurs new to AI turn their skills into real-world impact. Let’s go, Sangeeta! 🚀 Here's her product: https://equityresearch.ai/ -- P.S. If you’re looking for a supportive environment to learn and build with AI, the next cohort of the AI Builders Bootcamp starts this Friday (April 4th). 👉 https://lnkd.in/gPHZTdWs

AI gives us back our most valuable resource: time ⏳ For coders, the best way to tap into this is by using tools like ChatGPT, Cursor, or Co-pilot to help you write code faster. These tools have compressed my “weekend projects” into something I can knock out in an hour or less. In a recent YouTube video, I shared 5 such AI projects I’ve built in < 1 hour. There, I give step-by-step instructions for each and link example code in the video's description. 👉 Video link: https://lnkd.in/gRK76t5T

My AI Builders Newsletter just hit 1,000 subscribers! Starting a newsletter was something I avoided for a long time… My hesitation was that I didn’t want “yet another” place to post content (I was already posting on YouTube, Medium, LinkedIn, X, TikTok, and Instagram 😅). However, there is a fundamental difference between a newsletter and these other platforms. 👉 You have a direct line of communication to your audience. For example, of my 55,000 YouTube subscribers, 5-10% may watch a new upload. But with a newsletter, 30-50% of subscribers will open a new email! That’s why, in Jan 2025, I stopped posting on X, TikTok, and Instagram to focus more effort on my newsletter, which has made my content strategy clearer and more effective. If you’re building a personal brand or business... don’t procrastinate like me. Start a newsletter ASAP! -- P.S. If you’re trying to build real-world solutions with AI—join 1,000+ AI builders and get 30 free project ideas: https://lnkd.in/gf7z2eab

80% of people are NOT passionate about their job. Here’s the mental checklist I used to quit my job (to become an entrepreneur). 1) I want to work for myself No point in becoming self-employed if it’s not something you want for yourself. 2) I value freedom more than security The uncertainty of entreprenuership is probably what holds most people back. However, this uncertainty is the cost of freedom, which is a trade some are willing to make. 3) I value growth more than comfort This was the central reason I quit my job. I felt I was getting to comfortable in my FT role. While this might sound like a good thing, it often means you are leaving potential on the table. 4) Status (and other peoples’s opinions of me) aren’t more important to me than my goals Humans are hard-wired to care about status, which is another major deterrent for aspiring entrepreneurs. Lucky for me, I’ve sort of always been a weirdo, so I’m used to ignoring what people think of me :P 5) I had made money doing something that wasn’t a job (e.g. freelance and content creation) What gave me the confidence to think I could survive on my own was that I had previously generated income doing freelance work and content creation. This is why I think having a “side-hustle” is a critical prerequisite to quitting a job. 6) I have 1 year of runway in savings Realistically, in the first 1-2 years of entrepreneurship you will work 2X and make 1/2 as much as your FT role. Having runway gives you more time to build momentum and ultimately allows you to make bigger bets. -- What’s on your checklist? If you enjoyed this post, repost it and follow me (Shaw Talebi) for more.

After 15 months of failure and volatility, I finally matched my full-time income. I've realized what makes entrepreneurship hard is not coming up with ideas or learning new skills. Rather, it's that the road ahead is uncertain 😵‍💫 This is why I share my personal experience on socials–to (hopefully) make the path a little more clear and a little less painful for those who want to pursue something similar. In a recent blog post, I recap my journey to my first 15 months as an entrepreneur and (finally) matching my full-time income. 👉 Check it out here: https://lnkd.in/gpdixrcf

By the end of this post, you’ll be able to build an AI-powered resume optimization tool (in 6 steps) 🔨 Step 1: Create a markdown version of your resume with the desired format of your final resume (Note: ChatGPT can do this for you). Step 2: Read your markdown resume and job description into Python Step 3: Craft a prompt that combines your resume, the job description, and instructions for creating a new resume in Markdown. Step 4: Use ​OpenAI’s Python API​ to prompt GPT-4o-mini to rewrite your resume. Step 5: Convert the markdown output to HTML and then to PDF with the ​markdown​ and ​weasyprint​ libraries, respectively. Step 6: Create a simple web UI to make it easy to upload new resumes and optimize them for new JDs I've implemented this tool and shared the code for free on GitHub. Here's the link: https://lnkd.in/gw6wNKcn -- 👉 If you liked this post, repost it and follow me (Shaw Talebi) for more

Over the past 4 years, I've posted 100+ pieces of AI content on YouTube and Medium. While this all started as a way to structure my learning, it has evolved into much more than that today. Now, content is the primary way I get customers and makes up a significant portion of my revenue as a solopreneur. This Saturday (Mar 29) at 10:30AM CST, I'll share my entire content strategy and give away all my secrets. Join us live and ask me your content creation questions :) 👉 Register here: https://lu.ma/moqtb5p1

People who say "I don't have time to learn AI" are wrong. You can build your first AI project in less than 1 hour. Here are 5 LLM-based projects you can build fast. 1) YouTube to Blog Converter (Beginner) Extract YouTube transcripts with the youtube-transcript-api and use your favorite LLM API provider to translate it into a blog post. 2) PDF QA Chatbot (Beginner) Chat with sensitive documents locally by spinning up a chatbot using ollama and gradio. 3) RAG Chatbot (Intermediate) Use LlamaIndex to retrieve and synthesize information scattered across several PDFs. Bonus: use gradio to make a convenient UI. 4) Cluster Customer Feedback (Intermediate) Translate customer reviews into semantically meaningful vectors and categorize them using a traditional clustering algorithm (e.g. KMeans). 5) LinkedIn Post Writer (Advanced) Fine-tune GPT-4o-mini on your LinkedIn posts to have it generate new posts in your likeness. I share step-by-step instructions for each of these project ides (as well as example code) in a recent blog post. 👉 Check it out here: https://lnkd.in/gAP7x-A2

Today, I kicked off a new YouTube series on AI agents. The first video gives a beginner-friendly intro to agents (without the hype). Additionally, I share 3 common approaches for building agentic systems: LLM + Tools, LLM Workflows, and an LLM in a Loop. Future videos of these series will consist of my building a practical example project using each of these approaches. 👉 Check out the series intro here: https://lnkd.in/gk4TKZWA

Although APIs like OpenAI and Anthropic simplify production AI systems… … what if you need to self-host a custom LLM? Last week, Jeremy ARANCIO walked through a real-world case study where he did precisely this for one of his clients. Here's an overview of the project: - Gathered training data (input-output pairs of OCR spellcheck examples) - Fine-tuned several candidate LLMs - Deployed best model using vLLM and Docker - Used a batch process to apply the model to millions of examples ICYMI–the recording is now on The Data Entrepreneurs YouTube 👇 Check it out: https://lnkd.in/gHhEFg6h

In my first 12 months of entrepreneurship, I made 1/5th of my full-time salary. But that all changed when I made this simple mindset shift. When I quit my job to become an entrepreneur, I asked myself: "What work can I do to make a living?” Looking back, I realize this was the WRONG question. Instead, I should have asked: “What problems can I solve?” While the difference may seem subtle, these are fundamentally different mindsets. The first question is what I call the “employee mindset,” which focuses on the inputs you do that eventually generate value. The second is what I call the “entrepreneur mindset,” which focuses on the outcomes you want to create for your customer. In my first year of entrepreneurship, I had an employee’s mindset, which resulted in me working very, very hard just to make a fraction of my full-time income (basically minimum wage) 😅 It wasn’t until I adopted an “entrepreneur mindset” that I could match my full-time salary. So, if you are a new or aspiring entrepreneur, ask yourself… Am I focused more on doing a job or solving a specific pain point?