<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Talha Akhoon on Medium]]></title>
        <description><![CDATA[Stories by Talha Akhoon on Medium]]></description>
        <link>https://medium.com/@talha-akhoon?source=rss-6ed3d0233e13------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*5kAs-l1y8TlQ4v2DElY6ww.jpeg</url>
            <title>Stories by Talha Akhoon on Medium</title>
            <link>https://medium.com/@talha-akhoon?source=rss-6ed3d0233e13------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 06:48:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@talha-akhoon/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[YC Startup School Europe 2024: I Spent a Day With Paul Graham & 5 YC Founders — Here’s What They…]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://talha-akhoon.medium.com/what-paul-graham-and-europes-top-founders-really-said-at-yc-startup-school-europe-2024-6157e34ca35d?source=rss-6ed3d0233e13------2"><img src="https://cdn-images-1.medium.com/max/1260/1*StWtfIgxfA_UmftgenXsCQ.png" width="1260"></a></p><p class="medium-feed-snippet">Exclusive insights from Monzo, PostHog, Photoroom and high-growth YC founders&#x200A;&#x2014;&#x200A;and the &#xA3;1M advantage European startups have over Silicon&#x2026;</p><p class="medium-feed-link"><a href="https://talha-akhoon.medium.com/what-paul-graham-and-europes-top-founders-really-said-at-yc-startup-school-europe-2024-6157e34ca35d?source=rss-6ed3d0233e13------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://talha-akhoon.medium.com/what-paul-graham-and-europes-top-founders-really-said-at-yc-startup-school-europe-2024-6157e34ca35d?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/6157e34ca35d</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[ycombinator]]></category>
            <category><![CDATA[founders]]></category>
            <category><![CDATA[success]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Tue, 26 Nov 2024 10:01:18 GMT</pubDate>
            <atom:updated>2024-11-26T20:21:14.776Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Becoming the Ultimate SWE: Week 3 Roundup]]></title>
            <link>https://talha-akhoon.medium.com/becoming-the-ultimate-swe-week-3-roundup-797f6f427265?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/797f6f427265</guid>
            <category><![CDATA[public]]></category>
            <category><![CDATA[growth-mindset]]></category>
            <category><![CDATA[10x-engineer]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Wed, 28 Aug 2024 11:11:27 GMT</pubDate>
            <atom:updated>2024-08-28T11:18:17.882Z</atom:updated>
            <content:encoded><![CDATA[<h4>AI Caching Breakthrough, Demo Day Countdown, and the Road to 10x Engineering — Week 3 Unveiled!</h4><p>Welcome to my week 3 round-up of <a href="https://medium.com/@talha-akhoon/journey-to-becoming-the-ultimate-software-engineer-0c5a3739888a">my journey to become the ultimate software engineer</a>!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*_vaRjwxp8M9lYPUs.jpg" /></figure><h3>Monthly startup challenge</h3><p>Last week, I shared the gains I made with Vocall.ai - if you missed that update, see <a href="https://medium.com/@talha-akhoon/becoming-the-ultimate-swe-week-2-roundup-fd4ad4568cf0">Becoming the Ultimate SWE: Week 2 Roundup</a>.</p><h4>Progress</h4><p>Last week I managed to:</p><ol><li>Implement caching to avoid unnecessary requests to the LLM model. This was done by calculating a Jaccard similarity index. I’m still looking at other methods because I’m not convinced this is an accurate way of detecting whether something new has been said in the conversation.</li><li>Make some progress on adding a feedback loop. The aim is to allow the system to constantly “learn” from new experiences, without directly modifying the underlying LLM model itself.</li></ol><p>This week is the final week of this challenge. Next week is the first week of September.</p><p>What does that mean I hear you ask?<br><strong>DEMO DAY</strong></p><p>Next week, I aim to run through the app and provide a high level overview of the system (No sneak peeks this week — sorry!).</p><h3>LeetCode challenge progress</h3><h4>Progress:</h4><p>I solved 2 medium problems and 3 easy problems — total 5</p><ul><li><a href="https://leetcode.com/problems/design-linked-list/">707. Design Linked List</a></li><li><a href="https://leetcode.com/problems/design-browser-history/">1472. Design Browser History</a></li><li><a href="https://leetcode.com/problems/number-of-students-unable-to-eat-lunch/">1700. Number of Students Unable to Eat Lunch</a></li><li><a href="https://leetcode.com/problems/implement-stack-using-queues/">225. Implement Stack using Queues</a></li><li><a href="https://leetcode.com/problems/climbing-stairs/">70. Climbing Stairs</a></li></ul><p>I said I was going to solve 7 problems — clearly slacked 😢</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qGAFk76rvRsps2pM" /></figure><p>Yeah… Going to aim to do better this week!</p><h4>Goals for this week</h4><ul><li>Solve at least 7 problems — for definite!!!</li></ul><h3>Key learnings</h3><p>I did some reading around how to build a retrieval agent using LangChain in combination with a vector database like Pinecone.</p><p>The agent processes user queries by retrieving relevant context from the database and generating responses based on this information.</p><p>The approach enhances the accuracy and relevance of the responses, making it an effective tool for applications that require precise information retrieval integrated with the generative capabilities of language models.</p><p>I found this technique interesting because it significantly improves the quality of interactions by grounding responses in external knowledge.</p><h3>What are your goals for this week?</h3><p>Now that I’ve shared my goals for the upcoming week, I’m curious about yours! What are you working on to level up your software engineering skills? Are you tackling a new project, diving into a challenging algorithm, or exploring a cutting-edge technology?</p><p>Share your goals for the week in the comments below. Let’s motivate each other and create a community of aspiring 10x engineers. Who knows, your goal might inspire someone else, or you might find a coding buddy to keep you accountable!</p><h3>Stay Tuned and Connected:</h3><blockquote><em>Don’t miss a single step of this journey! Follow me here on Medium to get notified about each new update. Your engagement could be the push I need on a tough day, or the spark for your own growth. Let’s learn, grow, and code together.</em></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=797f6f427265" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Becoming the Ultimate SWE: Week 2 Roundup]]></title>
            <link>https://talha-akhoon.medium.com/becoming-the-ultimate-swe-week-2-roundup-fd4ad4568cf0?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd4ad4568cf0</guid>
            <category><![CDATA[10x-engineer]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[public]]></category>
            <category><![CDATA[growth-mindset]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Wed, 21 Aug 2024 09:01:42 GMT</pubDate>
            <atom:updated>2024-08-21T09:17:11.062Z</atom:updated>
            <content:encoded><![CDATA[<h4>Prompt Engineering Breakthroughs, Startup Progress, and LeetCode Wins — Week 2 Summary!</h4><p>Welcome to my week 2 round-up of <a href="https://medium.com/@talha-akhoon/journey-to-becoming-the-ultimate-software-engineer-0c5a3739888a">my journey to become the ultimate software engineer</a>!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vrhpVqaJorXf4pCG.jpg" /></figure><h3>Key learnings</h3><p>Last week I explored prompt engineering and alternative training options to fine-tuning an LLM model.</p><p>Prompt engineering plays a crucial role in improving the performance and reliability of LLM-based applications</p><p>Here are some techniques:</p><ul><li><strong>Role-based prompting:</strong> Assigning a specific role to the AI. It can help frame the context and perspective for responses.</li><li><strong>Instruction-based prompting:</strong> Giving clear, specific instructions. It ensures clear guidance for particular tasks.</li><li><strong>Few-shot learning:</strong> Providing examples within the prompt. It allows the model to learn from examples within the prompt itself.</li><li><strong>Chain-of-thought prompting:</strong> Guiding the model through step-by-step reasoning. It also encourages the model to provide more detailed, step-by-step reasoning.</li></ul><p>The above methods can be used in combination or individually, depending on the specific requirements of the task at hand.</p><h4>Goals for this week</h4><ul><li>Explore Retrieval Augmentation</li></ul><h3>Monthly startup challenge</h3><p>Last week, I mentioned I am currently working on a monthly startup challenge to build an MVP product, give a demo, and provide a high-level system design overview of how it all comes together — all within a month!</p><p>I am currently working on Vocall.ai — if you missed that update, see <a href="https://medium.talhaakhoon.dev/becoming-the-ultimate-swe-week-1-roundup-c1a003883113">Becoming the Ultimate SWE: Week 1 Roundup</a></p><h4>Progress</h4><p>Last week I managed to:</p><ul><li>Created a pipeline which allows for real-time extraction of presenting symptoms from the live transcript and suggests possible actions that the call-handler should take</li><li>Isolated the call-handing component and moved it into a modal</li><li>The modal is now able to present the transcriptions in real-time as well as the presenting symptoms and the list of actions. Confidence scores are also presented to enable the call-handler to make an informed decision.</li><li>As a bonus, I thought it would also be useful to suggest a list of questions that the call-handler could ask to prompt the patient for more details regarding their symptoms — this would also help the LLM model suggest actions with greater confidence and the call-handler to better decide on the next steps during the call.</li></ul><p>Here’s a little sneak peak…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5CFnYqMWm-mO77LHdT2kxw.png" /></figure><p>Let me know what you think of the progress so far.</p><h4>Goals for this week</h4><ul><li>Implement a caching strategy to avoid unnecessary requests to the LLM model</li><li>Explore Retrieval Augmentation to enrich prompt to LLM with relevant historical data.</li><li>Explore a more explicit feedback mechanism, possibly allowing the call handler to rate the effectiveness of the suggested actions as well as the outcome of each case</li></ul><h3>LeetCode challenge progress</h3><p>The villain in my storyline.</p><h4>Progress:</h4><p>I solved 4 easy problems and 1 medium problem (Yes, I know I said I was going to solve 5 easy problems):</p><ul><li><a href="https://leetcode.com/problems/concatenation-of-array/">1929. Concatenation of Array</a></li><li><a href="https://leetcode.com/problems/valid-parentheses/">20. Valid Parentheses</a></li><li><a href="https://leetcode.com/problems/min-stack/">155. Min Stack</a></li><li><a href="https://leetcode.com/problems/reverse-linked-list/">206. Reverse Linked List</a></li><li><a href="https://leetcode.com/problems/merge-two-sorted-lists/">21. Merge Two Sorted Lists</a></li></ul><h4>Goals for this week</h4><ul><li>Solve at least 7 problems</li></ul><h3>What are your goals for this week?</h3><p>Now that I’ve shared my goals for the upcoming week, I’m curious about yours! What are you working on to level up your software engineering skills? Are you tackling a new project, diving into a challenging algorithm, or exploring a cutting-edge technology?</p><p>Share your goals for the week in the comments below. Let’s motivate each other and create a community of aspiring 10x engineers. Who knows, your goal might inspire someone else, or you might find a coding buddy to keep you accountable!</p><h3>Stay Tuned and Connected:</h3><blockquote><em>Don’t miss a single step of this journey! Follow me here on Medium to get notified about each new update. Your engagement could be the push I need on a tough day, or the spark for your own growth. Let’s learn, grow, and code together.</em></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd4ad4568cf0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Leveraging LLMs: A Guide to Supercharging Your Applications with AI (Part 1 of 3)]]></title>
            <link>https://talha-akhoon.medium.com/leveraging-llms-a-guide-to-supercharging-your-applications-with-ai-part-1-of-3-bf46165f12cf?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf46165f12cf</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[supply-chain-management]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[large-language-models]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Fri, 16 Aug 2024 09:53:01 GMT</pubDate>
            <atom:updated>2024-08-17T20:56:11.265Z</atom:updated>
            <content:encoded><![CDATA[<h4>Is Your Supply Chain Stuck in the Stone Age? Here’s the AI Upgrade You Need Now</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*404c29-bchWYy2hZa1EM0g.png" /><figcaption>Your application after it has been supercharged with AI</figcaption></figure><p>In a recent article from my “Becoming the Ultimate SWE” (<a href="https://medium.talhaakhoon.dev/becoming-the-ultimate-swe-week-1-roundup-c1a003883113">Becoming the Ultimate SWE: Week 1 Roundup</a>) series, I briefly touched on how Large Language Models (LLMs) widens the possibilities by being able to continuously analyse and make the information digestible. Today, we’re expanding on that concept with a comprehensive look at integrating LLMs into your applications.</p><blockquote>Before we dive in, a quick note for new readers: I’m on a mission to become a 10x engineer, documenting my journey in the “<a href="https://medium.talhaakhoon.dev/journey-to-becoming-the-ultimate-software-engineer-0c5a3739888a">Becoming the Ultimate SWE</a>” series. It’s not just about my growth — it’s about pushing all of us to new heights. If you’ve got the fire in your belly to push harder, reach further — this series is for you to follow. See you there!</blockquote><h3>Context</h3><p>In today’s data-driven world, we face a paradox: an abundance of information, yet a struggle to utilise it fully. Approximately <strong>80%</strong> of data created worldwide is <strong>unstructured</strong> — think PDFs, spreadsheets, images, and videos. Surprisingly, organisations only tap into <strong>0.5%</strong> of this unstructured data [1].</p><p>The challenge? Traditional machine learning approaches are often too costly and complex for most organisations to implement effectively. This leaves a wealth of potential insights locked away in unstructured formats.</p><p>Large Language Models (LLMs) offer a solution. These AI tools, developed by companies like OpenAI, Anthropic, and Google, can analyse unstructured data at scale, making information more accessible. For most organisations, this means leveraging pre-trained models instead of building ML systems from scratch.</p><p>To adapt LLMs for specific business needs, two main approaches exist:</p><ol><li><strong>Fine-tuning</strong>: Powerful but expensive, requiring substantial data and potentially limiting future flexibility (you’re locked-in to using that model)</li><li><strong>Prompt engineering</strong>: Cost-effective and flexible, often yielding surprisingly good results.</li></ol><p>In this three-part series, we’ll explore increasingly advanced use cases of LLMs in business applications. Each part will focus on a specific scenario, demonstrating how LLMs can be integrated with other technologies to solve complex business problems.</p><h3>The Untapped Potential: LLMs in Supply Chain Management</h3><p>As a former Senior Software Engineer at Chill-Chain Ltd, a startup that developed innovative supply chain collaboration software, I’ve often reflected on how emerging technologies could transform the industry. While Chill-Chain unfortunately didn’t survive market pressures, the potential of our software — especially when combined with the power of Large Language Models (LLMs) — continues to inspire me. Let me show you how this could work in a critical scenario:</p><p><strong>Picture this:</strong> You’re an employee at Tesco PLC, working in the supply chain management department. It’s a typical Tuesday afternoon when suddenly, an urgent email lands in your inbox. The subject line makes your heart race: “URGENT: Potential contamination in organic spinach batches.”</p><p>As you read through the email, you realise the gravity of the situation. A supplier has detected possible E. coli contamination in several batches of organic spinach distributed to Tesco stores across the country. <strong>You need to act fast</strong> to:</p><ol><li><strong>Identify which specific products are affected</strong></li><li><strong>Determine which stores have received these batches</strong></li><li><strong>Alert the relevant store managers</strong> to <strong>remove the products</strong> from shelves</li><li><strong>Provide clear instructions for handling customer inquiries and potential returns</strong></li></ol><p>Time is of the essence. Every minute that passes could mean more contaminated products being sold to unsuspecting customers. But with hundreds of stores and thousands of products, manually processing this information and sending out alerts would take hours — time you simply don’t have.</p><p>This is where your newly implemented <strong>LLM-powered Product Safety Alert System</strong> comes into play. Within minutes of receiving the email:</p><ol><li>The system, powered by an LLM, immediately parses your email, extracting critical details about affected products, batch numbers, and contamination risks.</li><li>The platform automatically notifies and requests information from the relevant suppliers, using the LLM to generate clear, concise queries.</li><li>Leveraging Tesco’s extensive shipment data, the system swiftly identifies which stores received the potentially contaminated batches.</li><li>The LLM generates tailored alert messages for each affected store manager, distribution centre, and even delivery vehicles en route.</li><li>Using its understanding of Tesco’s supply chain operations, the LLM suggests optimal actions for product removal, replacement, and customer communication.</li><li>The system drafts press releases, social media updates, and customer service scripts, ensuring consistent and accurate information across all channels.</li></ol><p>As you watch the system work its magic, you breathe a sigh of relief. What could have been a logistical nightmare and potential PR disaster is now being handled swiftly and efficiently. Store managers are being notified, products are being pulled from shelves, and customer safety is being prioritised — all thanks to the power of LLMs and real-time data processing.</p><p>This vision of an LLM-enhanced supply chain management system isn’t just a theoretical concept. It’s a tangible possibility that could revolutionise how companies like Tesco handle critical situations, optimise their operations, and collaborate with partners.</p><p>Let’s explore how this LLM-powered system could work, and how you can implement similar solutions to supercharge your own supply chain operations in the age of AI.</p><h3>High Level Implementation Overview</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kuT0H2UTloQ6IEX3OHK6AA.png" /></figure><h4><strong>Email Alert Ingestion</strong>:</h4><ul><li>The system continuously monitors a designated email inbox for urgent alerts.</li><li>When an alert is received, it’s immediately passed to the LLM Email Parser.</li></ul><h4><strong>LLM Email Parser</strong>:</h4><ul><li>An LLM (e.g., GPT-3 or Claude) is fine-tuned to extract critical information from alert emails.</li><li>It identifies affected products, batch numbers, potential risks, and any other relevant details.</li><li>The extracted information is structured and sent to the product_safety_alerts Kafka stream.</li></ul><h4><strong>Kafka Streams</strong>:</h4><p>Three main Kafka streams are utilised:</p><ul><li><strong>product_safety_alerts</strong>: Contains structured data from the email alert</li><li><strong>product_info</strong>: Continuously updated stream of product information</li><li><strong>product_movements</strong>: Real-time data on product shipments and store inventories</li></ul><p>To generate:</p><ul><li><strong>product_safety_alerts_location: </strong>A stream processing application (e.g., using Kafka Streams API or Apache Flink) combines data from all three streams mentioned above to create this new stream with relevant information.</li></ul><h4><strong>LLM Message Composer</strong>:</h4><ul><li>Another LLM instance consumes input from the product_safety_alerts_location stream.</li><li>It generates tailored alert messages for different recipients (store managers, distribution centres, delivery vehicles).</li><li>The LLM also drafts customer communication materials, including press releases, social media updates, and customer service scripts.</li></ul><h4><strong>Alert Distribution System</strong>:</h4><ul><li>This system takes the composed messages and distributes them to the appropriate channels.</li><li>It ensures that each store manager, distribution centre, and affected delivery vehicle receives the relevant alert.</li></ul><h4><strong>Customer Communication</strong>:</h4><ul><li>The drafted customer-facing communications are sent to the appropriate departments for review and distribution.</li></ul><p>This implementation leverages the power of LLMs at two critical points:</p><ul><li>Parsing and extracting information from unstructured email alerts.</li><li>Composing tailored, context-aware messages for various stakeholders.</li></ul><p>By integrating LLMs with existing supply chain data and systems, we create a powerful, automated solution that can respond to critical situations with speed and precision that would be impossible to achieve manually.</p><h3>Conclusion: Embracing the LLM Revolution in Supply Chain Management</h3><p>The integration of Large Language Models (LLMs) into supply chain management systems represents a paradigm shift in handling critical situations and optimising operations. Key takeaways include:</p><ol><li><strong>Untapped Potential</strong>: LLMs offer a powerful solution to unlock insights from the 80% of unstructured data.</li><li><strong>Real-time Response</strong>: LLM-powered systems can dramatically reduce response times in crisis scenarios.</li><li><strong>Intelligent Automation</strong>: LLMs enable automation of complex, context-dependent tasks.</li><li><strong>Scalability and Precision</strong>: Combined with technologies like Kafka streams, LLMs allow for rapid, accurate handling of large-scale operations.</li><li><strong>Comprehensive Communication</strong>: LLMs ensure consistent, clear, and timely communication across all channels.</li></ol><p>The LLM-enhanced product safety alert system we’ve outlined is just the beginning. In the next part of this series, we’ll explore how LLMs can enhance customer support, generating responses enriched with business-specific context and balancing AI efficiency with human expertise.</p><p>As we embrace this AI revolution, it’s clear that LLMs are not just about keeping up with technology — they’re about reimagining what’s possible in operations and customer interactions. The time to start exploring these technologies is now.</p><blockquote>Stay tuned to learn how you can leverage these powerful AI tools to transform your business operations and drive innovation. Follow me here on Medium to get notified.</blockquote><h3>References</h3><p><strong>[1]</strong> <a href="https://www.deep-talk.ai/blog-posts/80-of-the-worlds-data-is-unstructured#:~:text=It%20is%20estimated%20that%20only,social%20networks%2C%20emails%2C%20etc">https://www.deep-talk.ai/blog-posts/80-of-the-worlds-data-is-unstructured#:~:text=It%20is%20estimated%20that%20only,social%20networks%2C%20emails%2C%20etc</a></p><h3>What area of your business do you think could benefit most from LLM integration?</h3><p>Share your thoughts in the comments below, and let’s discuss how AI could transform your specific use case!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf46165f12cf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Becoming the Ultimate SWE: Week 1 Roundup]]></title>
            <link>https://talha-akhoon.medium.com/becoming-the-ultimate-swe-week-1-roundup-c1a003883113?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1a003883113</guid>
            <category><![CDATA[growth-mindset]]></category>
            <category><![CDATA[public]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[10x-engineer]]></category>
            <category><![CDATA[career-development]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Tue, 13 Aug 2024 11:17:25 GMT</pubDate>
            <atom:updated>2024-08-15T14:43:13.460Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>TLDR:</strong> Busy first week; Exploring how LLMs can be integrated into applications, making significant strides on my first MVP for my monthly startup challenge, LeetCode practice — it’s been a packed seven days.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*hM1hSCa1HJngA3ZQ.png" /></figure><p>Welcome to the first instalment of <a href="https://medium.com/@talha-akhoon/journey-to-becoming-the-ultimate-software-engineer-0c5a3739888a">my journey to become the ultimate software engineer</a>!</p><p>Let’s break down what I’ve been up to, what I’ve learned, and the challenges I’ve faced. Whether you’re a fellow engineer looking to grow or just curious about the day-to-day of someone on a mission to reach 10x engineer status, I hope you’ll find something valuable in my experiences.</p><h3>Key learnings</h3><p>This week I explored how LLMs can be integrated into applications. Here’s a brief overview:</p><ul><li>Integrating LLMs &amp; other ML models allows for the opportunity to continuously analyse <strong>unstructured</strong> data and make the information accessible.</li><li>Around <strong>80%</strong> of data created in the world is estimated to be <strong>unstructured</strong>. This comes in the form of PDF documents, spreadsheets, images, audio, videos, etc. Organisations only utilise <strong>0.5%</strong> of all unstructured data <strong>[1]</strong>.</li><li>For most organisations traditional ML feature engineering is unthinkable due to costs</li><li>Instead, look at leveraging the inference capabilities of foundational models (ChatGPT, Claude, LLama, Google AI, Mosaic ML, Cohere, Mistral AI, ALEPH ALPHA)</li><li>Two options to make foundational models aware of business-specific knowledge: <strong>Option 1 — Fine Tuning, Option 2 — Prompt Engineering</strong></li><li><strong>Fine tuning</strong> a foundational model is expensive, requires substantial amount of data, and ties your organisation in to using the fine-tuned model in the future (a problem if you want to switch to a different model with enhanced capabilities)</li><li><strong>Prompt engineering </strong>is cost effective, flexible and can surprisingly give good results</li></ul><p>I’ll be diving deeper into this in another article this week.</p><p>Follow to get notified when it does drop.</p><h4>Goals for this week</h4><ul><li>Explore LLM use-cases for organisations</li></ul><h3>Monthly startup challenge</h3><p>To make things interesting, I’ve set myself a monthly startup challenge — 5 months (left of 2024), 5 MVPs. The goal is to build an MVP product, give a demo, and provide a high-level system design overview of how it all comes together — all within a month!</p><p>If there’s enough interest, I’ll consider open-sourcing some of my projects. (that’s one more reason for you to follow!)</p><h4>Idea</h4><p>Last week I started building Vocall.ai — An AI-assisted platform for healthcare services to allow call-handlers to make informed and practical decisions based off their conversations with patients over the phone.</p><h4>Progress</h4><p>Last week I managed to:</p><ul><li>Build the front-end UI to allow call-handlers to accept/reject incoming calls with Next.js and TailwindCSS</li><li>Build the core API and Websocket server with Golang</li><li>Integrate with Amazon Transcribe to return transcriptions generated from a call to the client (i.e call-handler) in real-time</li></ul><p>Here’s a little sneak peak…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gY01vgFUbf1U77eyuAAvUg.png" /></figure><h4>Goals for this week</h4><ul><li>Explore prompt engineering with LLMs to highlight symptoms and provide a confidence score against a set list of possible actions the call-handler should take</li></ul><h3>LeetCode challenge progress</h3><p>LeetCode for right now for me is the equivalent of leg day.<br>The equivalent of having to eat your veggies.<br>The equivalent of having to clean up after cooking up a delicious meal.</p><p>Why go through the pain?</p><blockquote>“There’s no better way to grow as a person than to do something you hate every day” — David Goggins</blockquote><p>That’s right! (<a href="https://www.youtube.com/watch?v=YBG7-zjFOdc">WHO’S GOING TO CARRY THE BOATS AND THE LOGS?!</a>)</p><p>The plan is to attempt a problem every day and move on once I’ve understood it.</p><h4>Progress:</h4><p>I solved 2 easy problems:</p><ul><li><a href="https://leetcode.com/problems/remove-duplicates-from-sorted-array/description/">26. Remove Duplicates from Sorted Array</a></li><li><a href="https://leetcode.com/problems/remove-element/">27. Remove Element</a></li></ul><h4>Goals for this week</h4><ul><li>Solve at least 5 easy problems</li></ul><h3>What are your goals for this week?</h3><p>Now that I’ve shared my goals for the upcoming week, I’m curious about yours! What are you working on to level up your software engineering skills? Are you tackling a new project, diving into a challenging algorithm, or exploring a cutting-edge technology?</p><p>Share your goals for the week in the comments below. Let’s motivate each other and create a community of aspiring 10x engineers. Who knows, your goal might inspire someone else, or you might find a coding buddy to keep you accountable!</p><h3>Stay Tuned and Connected:</h3><blockquote><em>Don’t miss a single step of this journey! Follow me here on Medium to get notified about each new update. Your engagement could be the push I need on a tough day, or the spark for your own growth. Let’s learn, grow, and code together.</em></blockquote><h3>References</h3><p><strong>[1]</strong> <a href="https://www.deep-talk.ai/blog-posts/80-of-the-worlds-data-is-unstructured#:~:text=It%20is%20estimated%20that%20only,social%20networks%2C%20emails%2C%20etc">https://www.deep-talk.ai/blog-posts/80-of-the-worlds-data-is-unstructured#:~:text=It%20is%20estimated%20that%20only,social%20networks%2C%20emails%2C%20etc</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1a003883113" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Journey to Becoming the Ultimate Software Engineer]]></title>
            <link>https://talha-akhoon.medium.com/journey-to-becoming-the-ultimate-software-engineer-0c5a3739888a?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/0c5a3739888a</guid>
            <category><![CDATA[public-learning]]></category>
            <category><![CDATA[career-development]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[growth-mindset]]></category>
            <category><![CDATA[10x-engineer]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Tue, 06 Aug 2024 09:46:44 GMT</pubDate>
            <atom:updated>2024-08-06T10:51:29.171Z</atom:updated>
            <content:encoded><![CDATA[<h4>And unlocking that 10x engineer skillset in the process</h4><p><strong>**NOT FOR THE FAINT-HEARTED**</strong></p><p>TLDR; I’m going to document my journey to becoming the ultimate software engineer — all while being humble 😜. They say growth stops when ego takes over — I’m aiming for infinite growth. Follow my journey, watch me make mistakes and maybe you’ll surpass me along the way. Challenge accepted? 😤</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/563/1*F7zEdrvUhaGmbiSnNNiRgg@2x.jpeg" /></figure><h3>Why Embark on This Journey? 🤔</h3><p>In the realm of software engineering, as in life, there’s always a higher peak to climb. I’ve been closely following the trajectories of top-tier SWEs who’ve scaled these peaks or are nearing the summit. Their journeys weren’t a walk in the park, but boy, did they earn their stripes.</p><p>My audacious goal? To become so <strong>exceptional</strong> that merely observing my work ethic makes you question your own. To inspire a fire in your belly to push harder, reach further.</p><h3>Why Go Public? 🤨</h3><p>If you’ve watched Naruto (if you haven’t then what are you doing with your life?!), then you’ll know that on his journey to becoming the leader of his village (given the title of “Hokage”), he inspired those around him and others that crossed his path (albeit for side quest/filler episodes).</p><h4>But what made him inspiring? 😮</h4><p>His indomitable spirit, sheer indefatigability and unwavering determination to become the best, even in the face of overwhelming adversity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/594/0*iA6zfU9CwgRjG0bV" /></figure><p>Whether it be those who struggled with self-doubt and those who believed they had already achieved success, both ends of the spectrum, felt a profound need to grow and become better when they encountered him constantly.</p><p>This created a beautiful symbiosis — Naruto gained worthy rivals, pushing him to new heights, while those around him were motivated to better themselves. It’s a virtuous cycle of growth I aim to replicate.</p><h3>The Battle Plan:</h3><ol><li>Attack everything. Retain what works, discard what doesn’t.</li><li>Refine, iterate, and repeat ad infinitum.</li><li>Share every nugget of wisdom, every hard-earned lesson, every eureka moment.</li><li>Leverage your feedback to course-correct and optimize my journey.</li></ol><h3>Official notice ‼️</h3><p>Here’s a quote:</p><blockquote>“By letting go of our egos and sharing our process, we allow for the possibility of people having an ongoing connection with us and our work, which helps us move more.” — <a href="https://amzn.to/3WQ3MQd">Show Your Work!</a> (affiliate link) by <a href="https://austinkleon.com/">Austin Kleon</a></blockquote><p>Consider this series of articles, my brag document. My promotion package (check out Hudeyfa Jama’s article: <a href="https://medium.com/@hudeyfa.jama/promotion-3b5ec8fe087f">Promotion</a>) in real-time. It’s my declaration of intent to outgrow, outlearn, and outperform — publicly and unapologetically.</p><h4>To my fellow competitive rivals,</h4><p><strong>This is my gauntlet thrown.</strong> <strong>My public commitment to becoming better than you.</strong> But here’s the twist — in doing so, I’m inviting you to become better than me.</p><p>Let’s push each other to unprecedented heights. <strong>Your move.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*22J0snCTpqPY6J6hn-Yp5A@2x.jpeg" /></figure><h4>Stay Tuned and Connected:</h4><p>Every week, I’ll be dropping a new update on Medium, chronicling my progress, sharing insights, and laying bare the challenges I face. While I’m not entirely sure if Medium is the optimal platform for this journey, it’s where we’ll start. If you have suggestions for better alternatives, I’m all ears!</p><blockquote>Don’t miss a single step of this journey! Follow me here on Medium to get notified about each new update. Your engagement could be the push I need on a tough day, or the spark for your own growth. Let’s learn, grow, and code together.</blockquote><p>Eagerly await your insights, critiques, and challenges. Until next time!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0c5a3739888a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Tough Decision: Development or Documentation?]]></title>
            <link>https://talha-akhoon.medium.com/when-circumstances-turn-your-team-lead-into-living-documentation-9b31bf911afc?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/9b31bf911afc</guid>
            <category><![CDATA[software-documentation]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[knowledge-management]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Tue, 16 Jul 2024 04:22:19 GMT</pubDate>
            <atom:updated>2024-07-22T17:58:51.666Z</atom:updated>
            <content:encoded><![CDATA[<p>Picture this (or not): You’re part of a high-pressure startup team — a tech lead. Customers need features, investors need targets to be met, and us… Well, we just need investment to make it to the next quarter so we can breathe. You need to make a decision, and you need to make it now — a decision that could affect the foreseeable future of the business…</p><h3><strong>Sacrifice development time to write documentation or give the order to go full steam ahead?</strong></h3><p>This wasn’t just a hypothetical scenario for me — it was my daily reality at a recent startup gig. Our dev team was small, scrappy, and racing against the clock. Our product? Promising. Our resources? Limited but highly skilled (shout out to my dev team!) Our to-do list? Never-ending.</p><p>In this pressure cooker environment, our tech lead became our human Wikipedia. Need to know why that function was named after someone’s cat? He knew. Curious about that cryptic comment from two years ago? He remembered. His brain held the master key to our digital kingdom.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*R1pxXU5y9NSmXITK.jpeg" /></figure><p>But here’s the thing: This wasn’t poor planning or negligence. It was survival.</p><p>With investors breathing down our necks and customers clamoring for features, every minute counted. Formal documentation often took a backseat to immediate development needs. Our lead’s encyclopedic knowledge became our secret weapon, allowing us to move at breakneck speed and tackle complex issues on the fly.</p><p>However, even secret weapons have their drawbacks:</p><ol><li>Information bottlenecks during crunch times (and let’s face it, it was always crunch time)</li><li>Our poor tech lead was constantly bombarded with questions</li><li>The terrifying “what if” of potentially onboarding new developers (if we ever got the funding)</li></ol><p>Despite our best efforts and coding marathons, our startup story didn’t have a happy ending. Economic headwinds proved too strong, and we ultimately faced liquidation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LDKkuN8DsEthNSvO.jpeg" /></figure><p>But the lessons we learned about balancing development and documentation in a high-pressure environment? Those are invaluable.</p><h3>So, how do you juggle knowledge sharing when you’re trying to keep the startup dream from sinking?</h3><p>In retrospect, our situation highlights how external pressures can shape internal practices. While comprehensive documentation is ideal, sometimes teams must adapt to their circumstances and rely on other methods of knowledge sharing.</p><p>For teams in similar situations, here are some lightweight approaches that helped us:</p><ol><li>Quick debriefing sessions after solving complex problems</li><li>Encouraging detailed commit messages and inline code comments</li><li>Regular, informal knowledge-sharing talks</li></ol><p>To team leads who find themselves as the primary knowledge holder: Your role is crucial, especially in challenging times. While you may not have time for extensive documentation, even small efforts to share and record information can be incredibly valuable.</p><h3>In short…</h3><p>Ultimately, every team’s situation is unique. While robust documentation is a <strong>worthy</strong> <strong>goal</strong>, it’s important to <strong>balance</strong> it with other <strong>pressing needs</strong>. The key is to find sustainable ways to share knowledge that fit within your team’s constraints and support your current objectives.</p><p>Best of luck!</p><h3>What are your tips on knowledge sharing in a startup environment?</h3><p>Now, I’m curious: What knowledge-sharing hacks have you discovered in your startup adventures? How do you balance the pressing need for development with the long-term benefits of documentation? Drop your wisdom in the comments below — who knows, your tip might just save another team from drowning in their own code!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9b31bf911afc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Navigating the Database Decision: Relational vs. NoSQL]]></title>
            <link>https://talha-akhoon.medium.com/navigating-the-database-decision-relational-vs-nosql-45d21bd7bc0d?source=rss-6ed3d0233e13------2</link>
            <guid isPermaLink="false">https://medium.com/p/45d21bd7bc0d</guid>
            <category><![CDATA[relational-databases]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[database-management]]></category>
            <category><![CDATA[data-architecture]]></category>
            <category><![CDATA[nosql]]></category>
            <dc:creator><![CDATA[Talha Akhoon]]></dc:creator>
            <pubDate>Thu, 11 Jul 2024 23:32:01 GMT</pubDate>
            <atom:updated>2024-08-27T13:31:19.650Z</atom:updated>
            <content:encoded><![CDATA[<h3>“To SQL or To NoSQL”</h3><h4>Deciding between Relational vs. NoSQL</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8lLW1jMuXt9kZ9vVv_EL1w.jpeg" /><figcaption>Image generated by DALL-E</figcaption></figure><p>You’re at a coffee shop, laptop open, trying to decide on a database for your new project. Your caffeine-fuelled brain is bouncing between SQL and NoSQL — I call it the coder’s block. Sound familiar?</p><p>I recently found myself in this exact situation, debating the database dilemma with a colleague over some questionably strong espresso. Here’s what I found:</p><blockquote><em>Choosing a database is a lot like picking the right coffee order — it all depends on what you need and when you need it.</em></blockquote><p>Let’s start with the classic espresso of databases:</p><h3>Relational</h3><p>You know, the classics like MySQL, PostgreSQL, and Oracle. They’re the perfectly structured shots that keep the data world running. When do you order this particular brew?</p><ul><li>When your data is organised and structured</li><li>When you need reliable transactions</li><li>When you need to combine data</li></ul><p>But sometimes, you need something a little more… flexible. Enter…</p><h3>NoSQL</h3><p>You’ve probably heard of MongoDB, Cassandra, and Redis . They come in different flavours:</p><ul><li>Document Stores: For when your data is as varied</li><li>Key-Value Stores: These are super fast and great for things like caching or session management.</li><li>Column-Family Stores: Cassandra falls into this category. They’re designed for high write and read throughput, perfect for big data applications.</li><li>Graph Databases: Neo4j is a popular one here. If you’re dealing with complex relationships (think social networks like Facebook), these are your go-to.</li></ul><h3>So, how do you choose?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/0*lZsvmX7OTGNsvH70.jpeg" /></figure><p>Well, it’s not always as simple as “I’ll have my usual”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/748/0*Kn87RWVvCRSKCgCD.jpg" /></figure><p>Sometimes you need the structured reliability of a classic espresso (relational). Other times, you need the flexibility of a create-your-own brew (NoSQL). And occasionally, you might even need both — a database blend, if you will.</p><p>Consider this:</p><ul><li>Is your data changing fast? NoSQL might be the one.</li><li>Need rock-solid consistency? Go with relational.</li><li>Expecting more users than a coffee shop on a Monday morning? NoSQL scales like nicely.</li><li>Don’t be afraid to mix and match. Use relational for your financials and NoSQL for your user profiles. It’s like pairing the right cake with your coffee.</li></ul><p>Remember, <strong>there’s no one-size-fits-all in databases</strong>. It’s about finding the right brew database for your specific needs.</p><p>So, take a deep breath, consider your options, and make your choice.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RwuaUnVm5L-A4Zkc.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=45d21bd7bc0d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>