Prithvi K Padmanabhan

Insightful session on how AI can bring benefits throughout different stages of the SDLC, including code explanation, generation, testing, code reviews, and monitoring in the production environment

10

1 Comment

𝐓𝐨𝐤𝐞𝐧𝐬 𝐢𝐧 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈 A 𝐭𝐨𝐤𝐞𝐧 is the 𝐬𝐦𝐚𝐥𝐥𝐞𝐬𝐭 𝐮𝐧𝐢𝐭 of text easily understood and processed by an AI model or LLM, often representing words, characters, or subwords. A token is essentially different from words i.e. a single word may comprise of multiple tokens. 𝐅𝐨𝐫 𝐞𝐱𝐚𝐦𝐩𝐥𝐞  • word[someone] -> token[some , one] • word[everyday] -> token[every , day] This process of breaking down text/words into tokens is called 𝐭𝐨𝐤𝐞𝐧𝐢𝐳𝐚𝐭𝐢𝐨𝐧, which helps the model understand and generate language. A popular tool provided by 𝐎𝐩𝐞𝐧𝐀𝐈 to know the number of tokens in a word can be found here : 𝐡𝐭𝐭𝐩𝐬://𝐩𝐥𝐚𝐭𝐟𝐨𝐫𝐦.𝐨𝐩𝐞𝐧𝐚𝐢.𝐜𝐨𝐦/𝐭𝐨𝐤𝐞𝐧𝐢𝐳𝐞𝐫 𝐁𝐞𝐧𝐞𝐟𝐢𝐭𝐬 𝐨𝐟 𝐓𝐨𝐤𝐞𝐧𝐬 During training, models learn relationships between tokens to generate coherent and contextually relevant responses. The way tokens are structured can impact the model's performance, with better tokenization leading to more accurate outputs. Models can optimize memory usage by focusing on tokens rather than entire sentences or phrases, leading to faster processing. 𝐁𝐢𝐥𝐥𝐢𝐧𝐠 Most of the large Language models are billed around the number of tokens it processed as input and the number of tokens generated as output. Total billed Tokens ~ Input tokens (as question + context) + Output tokens (generated Answer) 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐖𝐢𝐧𝐝𝐨𝐰 Another closely related concept is of Context Window, which refers to the maximum number of tokens (words, subwords, or characters) that a model can process at one time. This includes both input and output tokens combined. Generative models often have limits on the number of tokens they can handle in a single request (e.g., 4096 tokens). Some examples of Context window for certain models are given below:  𝐆𝐏𝐓-𝟑: The context window is up to 4096 tokens. 𝐆𝐏𝐓-𝟒: Offers different variants, with some configurations supporting a context window of up to 8192 tokens. 𝐁𝐄𝐑𝐓: Typically has a context window of 512 tokens. 𝐁𝐋𝐎𝐎𝐌: It has a context window of up to 2048 tokens. 𝐆𝐏𝐓-𝟒 𝐓𝐮𝐫𝐛𝐨 and 𝐋𝐨𝐧𝐠𝐟𝐨𝐫𝐦𝐞𝐫: They can handle context windows exceeding 8000 tokens. #GenerativeAI #MachineLearning

1

A few thoughts on #LLM usage in the small, prompted by my toying with #llama3 and #phi-3 on consumer hardware.

5

I am honored to receive a Certificate of Appreciation from senior director Amol Farkade, director Avinash Gatty, and my manager Rajeev Karnatak for my role in a significant project at GEP. We've successfully redesigned and launched the Spend Self-Serve application on GEP’s low code platform, Quantum. Key highlights of this project: 1. 𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦 𝐓𝐫𝐚𝐧𝐬𝐢𝐭𝐢𝐨𝐧: Rebuilt from scratch on Quantum, enhancing flexibility and speed. 2. 𝐂𝐮𝐬𝐭𝐨𝐦𝐞𝐫 𝐈𝐦𝐩𝐚𝐜𝐭: Successfully migrated 100+ customers to the new system, ensuring continuity and improved service. 3. 𝐓𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐎𝐯𝐞𝐫𝐡𝐚𝐮𝐥: Redesigned both the user interface and backend architecture, integrating MongoDB to enhance performance and scalability. 4. 𝐕𝐞𝐫𝐬𝐚𝐭𝐢𝐥𝐢𝐭𝐲: Expanded the application's capabilities to serve as a generic data processing tool, capable of handling various forms of Big Data. 5. 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐞𝐝 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭: Engineered the new version in under a year, a significant enhancement over the previous version which took two years. This achievement underscores our commitment to efficiency and innovation, leveraging GEP's Quantum platform to expedite the development process without compromising quality. This project was a massive team effort, and it’s gratifying to see our hard work pay off in such tangible ways. I'm excited about continuing to innovate and drive future projects forward. #technology #innovation #teamwork #GEP #Quantum #BigData #AI #Spend

122

18 Comments

amazing article ALSHAMA M S ! you outlined the essence of mentoring elegantly, showing how wisdom and skill flow seamlessly, if guidance and support is imparted delicately, patiently and responsibly! never knew stories assembled from real-world experiences could be so insightful! 🎉

12

1 Comment

Enterprises are building Domain specific small LLMs to make their Generative AI applications highly personalized to their use-cases and customer data! They need a cost effective, and scalable mechanism to host 100s of such models as part of what I call as "Gen AI OS". Here is our blog post on serving Gen AI models at scale using LoRA techniques on SageMaker. We show 2 design patterns on serving LoRA based LLMs at scale cost effectively. Blog: https://lnkd.in/giiubRE7 @GenerativeAI #AWS #LoRA #LLMs #SageMaker #S-LoRA #Punica #vLLM #MachineLearning #GenAI Venkatesh Krishnan, PhD Karan Jain Katherine Kwan Mehul Jain

124

2 Comments

[case study] An insurance company wanted to create #ML models to help gain insights based on historical application and claim data. SPR was up to this challenge, helping VGM leverage ML in their daily operations. #machinelearning #data #insurance #MLcasestudy https://lnkd.in/gDxMZ9Pb

1

I had read about the Zeigarnik effect and its application in products. However, it only recently dawned on me that it very well explains the calm and joy writing posts on my blog has given me over the years. Many professional and personal gurus have advocated writing to bust stress and drive clarity. This psychological explanation helps reinforce it. So, I wrote a blog post about it. What else? Check out link in the comments.

28

2 Comments

🚀 We're excited to share that Cendyn and Shiji Group are partnering to boost hotel profitability! 🙌👉 https://ow.ly/Ttzb30sCKLJ Our seamless integration between Cendyn CRS and Shiji Enterprise Platform PMS provides a unified experience to streamline operations and drive revenue growth. This collaboration empowers hoteliers to intelligently manage rates and availability, boost productivity and operational efficiency, enhance overall guest experiences, and implement highly targeted marketing strategies. By leveraging this integration, hoteliers can maximize profitability and create deeper connections with their guests. Don't miss out on the opportunity to empower your hotel business with smarter technology and stay ahead in the competitive hospitality market! #HospitalityTech #HotelManagement #HotelProfitability #RevenueOptimization #TravelInnovation

6

Ai : Hunanity's greatest creation for Good and Bad. To secure ourself #zerotrustai In the journey, MS brings up VASA-1 will AI model which essentially spins a selfie image and turn it into a talking clip of you. All you have to do is upload a photo along with a voice note and let the AI model do the talking for you.

2

Let’s face it, LLMs aren’t without limitations. In this discussion, Varun Grover shares the benefits of a RAG framework and provides an overview of how it works as well as practical use cases. Watch now: https://vrt.as/43MNQQx #AI #GenAI #LLM #RAG

2

*New episode of Level Up Prodcast with Malthi SS" In this episode, I chat with Sonali Nanda, who is Senior Vice President at Exxat. Prior to Exaat, Sonali was product leader with High Radius and managed a fairly large team with #Product, #Design, #Analytics and Ops. We spoke about: 💠 Managing different things as a Product Leader - people, product, strategy. 💠How to Foster Innovation and Creativity 💠Balancing Speedy Execution and Strategic Thinking 💠Advice for Women in Product Management 💠Managing Diverse Teams and Priorities Check the episode for tons of insights #ProductLeadership #WomenInLeadership #TechInnovation #CareerDevelopment #ProductManagement #EmpoweringWomen #LeadershipDevelopment #ProfessionalGrowth

23

2 Comments

Guidewire's Umang Jain and Yoganand Ghati recently collaborated on a Medium blog post, "Organizing Work for Simplicity and Improved Collaboration," to discuss the on-premise-to-cloud journey that Guidewire is making and how program management in its cloud platform engineering team has been instrumental in enabling teams and stakeholders to deliver consistently and predictably. #GuidewireCloud #cloudcomputing #insurance #innovation #insuranceinnovation #insurtech #insurancetechnology #propertyandcasualtyinsurance #development #engineering #software

4

The skillsoft course of Generative AI and its impact to business is very well defined and provided an insight of what Generative AI is and why it should be used by businesses to enhance efficiency, automate redundant tasks.

17

2 Comments

Early functional test prototype on dummy APIs An AI assistant that can order food, book cabs, buy groceries, fashion or electronics and more on one intuitive app that understands what you want and gives it to you directly - All for cheaper through ONDC Thankfully, our access to staging environment on ONDC is almost done, can’t wait to show a more functional prototype with the entire order flow! Would love to show a proper demo, sadly there’s a long way to go to make this functional, but really excited! #ONDC #chatgpt #startupindia

17

1 Comment

I have known Avni for almost ten years and I am super glad (and not surprised) that her customer-focused approach remains unchanged. We were fortunate to have Avni Savant speak to our teams at G2 India. The session was inspirational, and we learned a lot from her. Avni has a wealth of experience, and we are grateful for the valuable insights she shared with us. During the session brilliantly moderated by Aravind Raj, we discussed several topics and gained a few key takeaways: ⚡ Creating an environment that supports a "fail fast" mindset fosters innovation, and consistent leadership support is essential. Interestingly, this is something that Roger Valade aligns with strongly, and we are just in the middle of an innovation sprint. :) 🧠 We explored areas where AI can be implemented to solve product use cases. Avni stressed that we don't need to reinvent the wheel; instead, we should learn from companies that have successfully adopted AI. The most doable areas for AI implementation are to improve developer productivity and summarize insights and discoveries. 📊 We discussed how squad-specific goals (OKRs) can be integrated with the company's larger goals. For example, we can analyze if improving our product's onboarding experience impacts the adoption rate, and understand if the release was successful. Avni emphasized the importance of using data to make feature iterations and prioritization decisions. 💬 We discussed the significance of product development teams being close to the customers and leveraging customer interviews, and feedback from pre-sales, sales, and support teams. 💗 Lastly, Avni highlighted the importance of self-drivenness and having a clear point of view in one's career. She noted that this proactive approach is valued and appreciated. We are truly grateful to Avni for her time and lessons!

55

5 Comments