How to Improve Applications With Generative AI

Developing a 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 can quickly become overwhelming without a solid foundation. A messy structure leads to inefficiency, making scaling and collaboration difficult.  𝗪𝗵𝗲𝗿𝗲 𝗦𝗵𝗼𝘂𝗹𝗱 𝗬𝗼𝘂 𝗕𝗲𝗴𝗶𝗻?   To streamline development, I’ve designed a 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 that prioritizes 𝘀𝗰𝗮𝗹𝗮𝗯𝗶𝗹𝗶𝘁𝘆, 𝗺𝗮𝗶𝗻𝘁𝗮𝗶𝗻𝗮𝗯𝗶𝗹𝗶𝘁𝘆, 𝗮𝗻𝗱 𝗰𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻.  𝗞𝗲𝘆 𝗖𝗼𝗺𝗽𝗼𝗻𝗲𝗻𝘁𝘀 𝗼𝗳 𝘁𝗵𝗲 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲  ✅ 𝗰𝗼𝗻𝗳𝗶𝗴/ – YAML-based configurations to separate settings from code.   ✅ 𝘀𝗿𝗰/ – Modularized core logic, including 𝗹𝗹𝗺/ and 𝗽𝗿𝗼𝗺𝗽𝘁_𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴/ components.   ✅ 𝗱𝗮𝘁𝗮/ – Organized storage for embeddings, prompts, and datasets.   ✅ 𝗲𝘅𝗮𝗺𝗽𝗹𝗲𝘀/ – Ready-to-use scripts for real-world use cases (e.g., chat sessions, prompt chaining).   ✅ 𝗻𝗼𝘁𝗲𝗯𝗼𝗼𝗸𝘀/ – Jupyter notebooks for rapid experimentation and analysis.  𝗕𝗲𝘀𝘁 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗲𝘀 𝗳𝗼𝗿 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁  🔹 Use 𝗬𝗔𝗠𝗟 for clean, readable configurations.   🔹 Implement 𝗲𝗿𝗿𝗼𝗿 𝗵𝗮𝗻𝗱𝗹𝗶𝗻𝗴 & 𝗹𝗼𝗴𝗴𝗶𝗻𝗴 for efficient debugging.   🔹 Apply 𝗿𝗮𝘁𝗲 𝗹𝗶𝗺𝗶𝘁𝗶𝗻𝗴 to manage API consumption effectively.   🔹 Maintain a 𝗰𝗹𝗲𝗮𝗿 𝘀𝗲𝗽𝗮𝗿𝗮𝘁𝗶𝗼𝗻 𝗼𝗳 𝗺𝗼𝗱𝗲𝗹 𝗰𝗹𝗶𝗲𝗻𝘁𝘀 for flexibility.   🔹 Optimize performance through 𝘀𝗺𝗮𝗿𝘁 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 𝗰𝗮𝗰𝗵𝗶𝗻𝗴.   🔹 𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 to ensure seamless team collaboration.   🔹 Leverage 𝗝𝘂𝗽𝘆𝘁𝗲𝗿 𝗻𝗼𝘁𝗲𝗯𝗼𝗼𝗸𝘀 for quick experimentation before production deployment.  𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝗦𝘁𝗮𝗿𝘁𝗲𝗱  • Clone the repository & install dependencies.   • Configure your model using the provided YAML files (**config/**).   • Explore 𝗲𝘅𝗮𝗺𝗽𝗹𝗲𝘀/ for real-world implementations.   • Utilize 𝗝𝘂𝗽𝘆𝘁𝗲𝗿 𝗻𝗼𝘁𝗲𝗯𝗼𝗼𝗸𝘀 for fine-tuning and testing.      𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿 𝗧𝗶𝗽𝘀   ✔ Follow 𝗺𝗼𝗱𝘂𝗹𝗮𝗿 𝗱𝗲𝘀𝗶𝗴𝗻 𝗽𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲𝘀 to keep your codebase clean.   ✔ Write 𝘂𝗻𝗶𝘁 𝘁𝗲𝘀𝘁𝘀 for new components to ensure reliability.   ✔ Monitor 𝘁𝗼𝗸𝗲𝗻 𝘂𝘀𝗮𝗴𝗲 & 𝗔𝗣𝗜 𝗹𝗶𝗺𝗶𝘁𝘀 to optimize costs.   ✔ Keep 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝘂𝗽𝗱𝗮𝘁𝗲𝗱 for easy scalability.  By adopting this structured approach, you can 𝗳𝗼𝗰𝘂𝘀 𝗼𝗻 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗶𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝘄𝗿𝗲𝘀𝘁𝗹𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗽𝗿𝗼𝗷𝗲𝗰𝘁 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻.  How do you structure your 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 projects? Share your thoughts in the comments!  

Exactly a year ago, we embarked on a transformative journey in application modernization, specifically harnessing generative AI to overhaul one of our client’s legacy systems. This initiative was challenging yet crucial for staying competitive: - Migrating outdated codebases - Mitigating high manual coding costs - Integrating legacy systems with cutting-edge platforms - Aligning technological upgrades with strategic business objectives Reflecting on this journey, here are the key lessons and outcomes we achieved through Gen AI in application modernization: [1] Assess Application Portfolio. We started by analyzing which applications were both outdated and critical, identifying those with the highest ROI for modernization.  This targeted approach helped prioritize efforts effectively. [2] Prioritize Practical Use Cases for Generative AI. For instance, automating code conversion from COBOL to Java reduced the overall manual coding time by 60%, significantly decreasing costs and increasing efficiency. [3] Pilot Gen AI Projects. We piloted a well-defined module, leading to a 30% reduction in time-to-market for new features, translating into faster responses to market demands and improved customer satisfaction. [4] Communicate Success and Scale Gradually. Post-pilot, we tracked key metrics such as code review time, deployment bugs, and overall time saved, demonstrating substantial business impacts to stakeholders and securing buy-in for wider implementation. [5] Embrace Change Management. We treated AI integration as a critical change in the operational model, aligning processes and stakeholder expectations with new technological capabilities. [6] Utilize Automation to Drive Innovation. Leveraging AI for routine coding tasks not only freed up developer time for strategic projects but also improved code quality by over 40%, reducing bugs and vulnerabilities significantly. [7] Opt for Managed Services When Appropriate. Managed services for routine maintenance allowed us to reallocate resources towards innovative projects, further driving our strategic objectives. Bonus Point: Establish a Center of Excellence (CoE). We have established CoE within our organization. It spearheaded AI implementations and established governance models, setting a benchmark for best practices that accelerated our learning curve and minimized pitfalls. You could modernize your legacy app by following similar steps! #modernization #appmodernization #legacysystem #genai #simform — PS. Visit my profile, Hiren Dhaduk, & subscribe to my weekly newsletter: - Get product engineering insights. - Catch up on the latest software trends. - Discover successful development strategies.

Most people want to use Generative AI. Fewer know how to build it. Even fewer know how to build it right. That’s where a roadmap like this becomes essential. I just went through this detailed Generative AI Roadmap, and it lays out a learning path from fundamentals all the way to deploying AI agents and real-world apps. If you're serious about building GenAI skills, here’s what’s included: - Start with core concepts: supervised vs. unsupervised learning, overfitting, basic Python, matrix ops, probability - Move into generative modeling: RNNs, autoencoders, latent space, backprop, VAEs - Deep dive into GANs & diffusion models: StyleGAN, CycleGAN, Stable Diffusion, U-Nets - Explore LLMs for text generation: transformers, attention, prompt engineering, few-shot learning - Go beyond text: music, audio, synthetic data, 3D generation - Learn fine-tuning techniques: LoRA, PEFT, instruction tuning - Then get hands-on with deployment: containerization, quantization, APIs, scaling - And finally, build AI agents with LangChain, CrewAI, and n8n—tying perception, reasoning, and action into workflows This roadmap is perfect for developers, ML engineers, and even product teams looking to understand what it really takes to go from an idea to a working GenAI app. -- Join our Newsletter with 137K Subscribers — www.theravitshow.com

Creating a Proof of Concept Generative AI app is deceptively easy. In a few hours, you can hack together a prototype that meets 60% of your requirements using a framework like LangChain. However, managing expectations around the finished product's timelines is crucial. The complexity increases exponentially beyond the initial phase. Achieving 90% of your product requirements demands rigorous test-driven development to expose and protect against risks. Sourcing for a wide variety of real user questions (both in-domain and out-of-domain), setting the temperature to 0, and using the seed parameter in the OpenAI API (https://lnkd.in/gahRjUKr) can significantly enhance the predictability and testability of your code changes. Addressing the final 10% is the most challenging. It rarely involves direct solutions. Instead, it's about identifying and blocking undesirable interactions. Each Large Language Model (LLM) has its peculiarities, presenting unpredictable and non-repeatable edge cases. These issues range from protection against prompt injections to difficulties maintaining domain restrictions. I've previously outlined some strategies to tackle these challenges in a detailed post: https://lnkd.in/gjFTMpbR A strong collaboration between technology and business teams is essential for developing these systems successfully. We engineers must adopt a product-centric mindset, communicate openly, and stay agile to adopt best practices as they continuously evolve. Everyone in this field is learning as we go, and I'll continue to share my insights on building production and customer-facing applications as I discover them. Follow me for more tips on building successful ML and LLM products! Medium: https://lnkd.in/g2jAJn5 X: https://lnkd.in/g_JbKEkM #generativeai #llm #nlp #artificialintelligence #mlops #llmops

There's a lot of great entry points into building generative #AI applications, but even after you get to that first "aha" moment, there's a ton of work left to do to get to production. Just like with regular software applications, beyond building the POC, there's things like: - creating fallbacks - checking latency - adding guardrails - call tracing - output testing On top of that, #GenAI also tends to hallucinate, so you've got to add in separate evaluation metrics to check if your application is putting out the right answers, and not just making things up. My favorite tools from moving from that first step towards more production use cases are: - Portkey for creating an AI gateway that allows me to set up fallbacks and guardrails - Arize AI for tracing calls, doing hallucination checks, and evals - NVIDIA for hosting and inference What are your favorite tools?

🧠 Generative AI Frameworks & Tools: A Guide to Key Players in the Ecosystem Generative AI is revolutionizing the way applications are built, and frameworks/tools play a critical role in streamlining this process. Let’s break down some of the top frameworks shaping the generative AI landscape: 🔹 LangChain Purpose: A general-purpose framework designed to integrate multiple Large Language Models (LLMs) and external tools. Use Case: Ideal for building conversational agents, custom workflows, and task automation with LLMs. 🔹 LlamaIndex Purpose: Built for Retrieval-Augmented Generation (RAG) workflows, enabling production-ready applications. Core Features:Efficient indexing of structured and unstructured data. Seamless integration with LLMs for intelligent data retrieval. Use Case: Powering document-based question answering and search applications. 🔹 Haystack Purpose: A robust framework for creating modern, search-based pipelines with LLMs. Core Features:Vector search for semantic similarity. Scalable pipelines for large datasets. Use Case: Useful for semantic search, knowledge base querying, and building intelligent search engines. 🔹 Hugging Face Purpose: A centralized hub for pre-trained models and datasets, as well as tools for fine-tuning. Core Features:Model hosting, versioning, and deployment. Community-contributed datasets and transformers for various NLP tasks. Use Case: Fine-tuning models for text generation, classification, or summarization tasks. Why These Frameworks Matter These tools simplify the integration of LLMs into real-world applications, reduce development time, and enhance capabilities like contextual search, semantic understanding, and generative workflows. 💡 Generative AI is no longer just about building; it’s about building smarter and faster. Which framework are you leveraging for your projects? Let’s discuss how these tools are shaping the future of AI development.

Building GenAI applications today feels like full-stack dev in the early 2000s. Fragmented, evolving, and exhilarating. From Multi Modals to Fine-tuning you’re not just coding, you’re composing an experience. So, what powers real-world GenAI Applications? It's the ecosystem wrapped around it. 𝟏. 𝐌𝐨𝐝𝐞𝐥𝐬: Pick your engine: GPT-4, Claude, Mistral, DeepSeek, Gemini 2.5, Llama. The brain. 𝟐. 𝐂𝐨𝐝𝐞 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤𝐬: LangChain, CrewAI, LangGraph, Google-ADK, OpenAI SDK, Hugging Face Transformers. Build agents, memory, and workflows. 𝟑. 𝐍𝐨 𝐂𝐨𝐝𝐞 𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦: Langflow, Flowise, n8n. Drag-and-drop for fast prototyping. 𝟒. 𝐓𝐨𝐨𝐥 𝐂𝐚𝐥𝐥𝐢𝐧𝐠: OpenAI Function Calling, Anthropic MCP, LangChain Tools, Agents API, Gorilla LLM Router. Execute code, fetch data, or trigger actions live. 𝟓. 𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬: LlamaGuard, Rebuff, Guardrails AI or Giskard. Keep things safe, brand-aligned, and compliant. 𝟔. 𝐓𝐞𝐬𝐭𝐢𝐧𝐠 & 𝐄𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐨𝐧: LangSmith, TruLens, Helicone or Ragas. Track quality, spot drift, test responses. 𝟕. 𝐃𝐞𝐩𝐥𝐨𝐲𝐦𝐞𝐧𝐭: Use Amazon Bedrock, Google Vertex AI, Azure AI Foundary or BentoML. 𝟖. 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐛𝐢𝐥𝐢𝐭𝐲: Watch prompt flows, latency, and usage with LangSmith, PromptLayer, Helicone or WhyLabs. 𝐁𝐨𝐭𝐭𝐨𝐦 𝐥𝐢𝐧𝐞: The model is just one piece. To make your GenAI applications work seamlessly in the real world, you need scalable architecture, safety, and iteration built in. 👉 Ready to build the next-gen GenAI app? Let’s discuss how to scale and refine your vision. #GenerativeAI #AIApplications #ArtificialIntelligence #MachineLearning #AIFrameworks #TechEcosystem

From Meh to Mind-Blowing: A Kano Model Hack for Generative AI Let's talk about something I've been thinking about lately: how generative AI (Gen AI) can transform businesses and why the Kano Model is the perfect lens to prioritize its adoption. Gen AI isn't new, but its explosion into the mainstream (think ChatGPT, Gemini) has turned it into a game-changer. The real question isn't if to use it, but how to use it strategically. Here's how the Kano Model can guide your approach: 1️⃣ Start with the Basics: "Must-Have" AI Today, simply using Gen AI tools is becoming a baseline expectation. Customers already assume you're leveraging these tools for faster responses, content creation, or data analysis. If you're not here yet, you're already playing catch-up. 2️⃣ Level Up: "Performance-Driven" AI This is where you stand out. By tailoring Gen AI to your business feeding it your data, refining outputs for your audience, or integrating it into workflows you turn a generic tool into a competitive edge. Think smarter chatbots, hyper-relevant marketing, or real-time analytics. 3️⃣ The Magic Moment: "Delightful" AI Here's where you surprise people. Imagine AI that anticipates needs before customers ask, adapts in real-time based on behavior, or creates entirely new experiences. Think self-improving systems or creative solutions that redefine what's possible. This isn't just "innovation" it's future-proofing. Why This Matters Gen AI isn't a trend it's a tidal wave. Companies that treat it as a checkbox or wait for others to innovate ("We use ChatGPT!") will stagnate. Those who reimagine processes, products, and customer journeys around AI will lead their industries. The risk? Waiting too long. Early adopters aren't just gaining efficiency they're shaping expectations. Falling behind could mean playing an endless game of catch-up. My Challenge to You Start small, but think big. Master the basics, then aim for differentiation. And always ask: "How could AI not just meet but redefine what's possible here?" I've seen firsthand how this framework drives real impact. What do you think? Could the Kano Model shape your AI strategy? Let's chat in the comments! 👇 (P.S. If you're stuck at "Where do I even start?", let me know happy to share practical steps)

Free 𝗧𝗲𝗺𝗽𝗹𝗮𝘁𝗲 𝗳𝗼𝗿 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗻𝗱 𝗘𝘃𝗼𝗹𝘃𝗶𝗻𝗴 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 - steal it with pride! I have been developing Agentic Systems for around two years now. The same patterns keep emerging. Today, I am sharing my system of how to approach development of LLM based applications from idea to production. Let’s zoom in: 𝟭. Define a problem you want to solve: is GenAI even needed? 𝟮. Build a Prototype: figure out if the solution is feasible. 𝟯. Define Performance Metrics: you must have output metrics defined for how you will measure success of your application. 𝟰. Define Evals: split the above into smaller input metrics that can move the key metrics forward. Decompose them into tasks that could be automated and move the given input metrics. Define Evals for each. Store the Evals in your Observability Platform. ℹ️ Steps 𝟭. - 𝟰. are where AI Product Managers can help, but can also be handled by AI Engineers. 𝟱. Build a PoC: it can be simple (excel sheet) or more complex (user facing UI). Regardless of what it is, expose it to the users for feedback as soon as possible. 𝟲. Instrument your application: gather traces and human feedback and store it in an Observability Platform next to previously stored Evals. 𝟳. Run Evals on traced data: traces contain inputs and outputs of your application, run evals on top of them. 𝟴. Analyse Failing Evals and negative user feedback: this data is gold as it specifically pinpoints where the Agentic System needs improvement. 𝟵. Use data from the previous step to improve your application - prompt engineer, improve AI system topology, finetune models etc. Make sure that the changes move Evals into the right direction. 𝟭𝟬. Build and expose the improved application to the users. 𝟭𝟭. Monitor the application in production: this comes out of the box - you have implemented evaluations and traces for development purposes, they can be reused for monitoring. Configure specific alerting thresholds and enjoy the peace of mind. ✅ 𝗖𝗼𝗻𝘁𝗶𝗻𝘂𝗼𝘂𝘀 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻: ➡️ Run steps 𝟲. - 𝟭𝟬. to continuously improve and evolve your application. ➡️ As you build up in complexity, new requirements can be added to the same application, this includes running steps 𝟭. - 𝟱. and attaching the new logic as routes to your Agentic System. ➡️ You start off with a simple Chatbot and add a route that can classify user queries to take action (e.g. add items to a shopping cart). I will be teaching how to apply this system hands-on and in detail as part of End-to-End AI Engineering Bootcamp (𝟭𝟬% 𝗱𝗶𝘀𝗰𝗼𝘂𝗻𝘁 𝗰𝗼𝗱𝗲: Kickoff10 ): https://lnkd.in/dGVhxAD9 What is your experience in evolving Agentic Systems? Let me know in the comments 👇 #LLM #AI #MachineLearning

Get what you want out of Generative AI by being intentional about the inputs you provide. Some practical tips: 1. Define the Specific Output. Don't ask for "a marketing email." Ask for "a 200-word email to SaaS founders announcing a new integration, written in a conversational tone with one clear call-to-action button." 2. Provide Complete Context. "I'm managing a 3-month website redesign for a healthcare client. Team of 4 people. Budget is $50K. Client wants to launch before their conference in December. Create a project timeline with weekly milestones." 3. Specify the Format and Constraints. Instead of hoping for the right format, tell it exactly what you need: "Create this as a bulleted action plan with deadlines, assigned team members, and budget allocations for each phase." 4. Build in Next Steps. End every prompt with what comes next: "After you create this timeline, I'll need you to identify the 3 biggest risk factors and suggest mitigation strategies." Being specific from the start allows you to avoid an extended back-and-forth. #legaltech #innovation #law #business #learning