Akhilesh K.

Senior developers can stay connected to the codebase, provide productive hands-on experience, maintain updated technical skills, and remain highly productive even without coding regularly. However, over time, not writing code or delivering key features can lead to a sense of moral high ground because they are not involved in meeting strict customer deadlines. On a personal note, I get little to no time to write code, but I ensure to contribute to at least one project per quarter, just for the love of it. It is one of the few job profiles that allows individual contributors to operate at such top levels. Making the most of it. Regardless, interesting read.

12

🚀 Bengaluru-based Conversational AI Startup Gnani AI Raises $3M 💰 Gnani AI, India's first voice-first small language model (SLM) for vernacular languages, has secured $3M in funding from tech holding company Info Edge. The investment will fuel the startup's expansion plans. Key Highlights: • Gnani AI is building a series of voice-based SLMs, trained on millions of hours of proprietary audio datasets and billions of conversations in 12 Indian languages • The startup has over 100 customers in banking, financial services, insurance, telecom, automotive, and healthcare industries • Gnani AI's models run on the edge, reducing costs and enabling quick fine-tuning based on enterprise needs • The startup specializes in creating customized models for individual companies, adding real value to their data CEO Ganesh Gopalan emphasizes the importance of multimodal models, focusing on voice and text fusion. By taking their models to enterprises and helping them build on top of it, Gnani AI is revolutionizing the conversational AI landscape in India. #ConversationalAI #VoiceFirst #IndianLanguages #Funding #Expansion

12

1 Comment

India Rank 1, Global Rank 90 in #Atcoder #Beginner #352 Instead of stream, I will write my solution ideas here. I ended up solving ABCDE and G. I could not debuging WA on 2 tests for F. Here is what I did for all the problems. A: If X < Y, then we will take inbound train, start at station X and end at station Y, so station Z must be between X and Y. X<=Z<=Y must hold. If X >= Y, then we will take outbound train, start at station Y and end at station X, so station Z must be between X and Y. Y<=Z<=X must hold. B: We are given that S is subsequence of T, we should print all the index using which we can form S. Also, notice that indirectly guarantees that there will be a unique set of indexes. We can use greedy to find set of thoses indexses. This is exactly same problem B of previous codeforces round (namely round 943) C: Head of Pn is at height sum(Ai) + B_{pn} - A_{pn} So, we will have to find maximum value of B_{pn}-A_{pn), and print sum(Ai)+mx D: We can use sliding window technique here. For each i, we will maintain set of index where [i,i+k-1] exists. Then check difference of minimum and maximum value in the set. When we want to find new set for j=i+1, we erase position of p[i] and add p[j+k] E: Eaach clique gives us K*(K-1)/2 edges, and most of them are redundant, so we should ignore them. For each clique we will add K-1 edges in our graph, Ai,1 -> Ai,j with weight Ci for each 2<=j<=K_i. If graph is disconnected we should print -1, otherwise we can using any MST finding algorithm to MST of this graph. G - Lets first try a bruteforce idea. Let p_j be the probability that we have select j socks and all of them are unqiue? Then answer is 1+sum(p_j) 1<=j<=n Let tot = sim(Ai), ie total no of socks. To find p_j, let the socks we have after j selects is set c1,c2,...,cj If ck is selected at kth sock 1<=k<=j, then its probability is A_{ck}/(tot-k+1), denominator is no of socks remaining after selecting first k-1 socks. Now notice that value of numerator and denominator is same for this set irrespective of the order of selection. Numerator is prod(A_{ck)} and denominator is prod(tot-k+1) 1<k<=j So we can fix the set, find prod of all values of A_i in selected set, and then arrange them using j! ways. Let F(x) = prod(1+Ai*x) = fj*x^j Then this fj is sum of all possible value of numerators. p_j = fj*j!/(required denominator) We can use divide and conquer with NTT to find F. Its similar to build function for segment tree. Following is the pseudo code build(l,r){ if(l==r) return {1+al} mid=(l+r)/2 return convolution(build(l,mid),build(mid+1,r)) } F - I passed 60/62 tests for this problem. I could not debug corner testcase. I will write my idea in comments because I have exhaused char limit for a linkedin post.

610

23 Comments

“My worst technical interview experience” . While trying to switch, I interviewed at Coinbase for the role of SDE 2. Now, let me clarify, the issue wasn’t with Coinbase—it was my incorrect expectations and irrelevant preparation. I assumed it would be a standard DSA round, but it was a Q coding round. I didn’t clarify with the recruiter and wasted time optimizing a straightforward problem. The real test was applying OOP concepts and refactoring. 🔹 Why was this so bad → I could have done well if I had understood the interview expectations and not been stuck in the "Leetcode mindset." 🔹 What I learned → Always clarify what concepts will be tested in the interview and read about others' experiences to prepare. However, finding detailed interview experiences isn’t easy. Many are vague, and don’t cover the specific challenges you face. That’s why it’s better to prepare directly with senior engineers from your target companies. You can connect with them on a FREE session here: https://lnkd.in/g6HGy4Dd From sharing what questions to prepare to conducting mock interviews and providing feedback to boost your confidence, these experts at Preplaced will help you with everything. 💫 #interviewexperience #coinbase #collab #preplacedcollab #interviewquestions

203

12 Comments

Few people join as a SDE-1 and within a span of ~7 years, they promote to a very senior level(L6 equivalent to Staff Role) in a company whereas same achievement is hard to realize when you're switching companies. So, what makes these employees so special? A few points: - They joined these companies as SDE-1 - Implemented projects end to end - very pro-active in all things: reviews/discussions/coding. - highly biased for taking actions (critical skillset for growth) - were lucky enough to some extent in getting good projects - were hard working enough to find optimizations & propose as new projects in the product - filled with lots of motivation & drive: showing true ownership at the deepest level one could possibly imagine Staying in a company helps you build context over the years, which the lateral hires within few months might or might not be able to grasp. This post is just a reminder: Switching would probably make money but staying in a company with a little bit of luck can do wonders(both money & role). PS: I did 2 switches in my career and I can relate with this post.

550

12 Comments

Penned this blogpost on the evolution of learning to rank models in Autosuggest at Flipkart Search. Do give it a read! https://lnkd.in/g6x7R67q

94

6 Comments

“The collision is not there anymore, Google, but thanks for almost getting me into one”! While I understand the intent behind it, sometimes, it doesn’t make sense to prompt a user (multiple times) with a call to action while driving actively- especially on Indian highways. Not to mention that almost 30% of the space on the screen is covered with the prompt. In the fight between being minimal and being a feature factory, sometimes the latter wins. Maybe GMaps could take a lesson in UI/UX from Gpay :) That is why, it should be mandatory for product managers and designers to dogfood their products. Use a “stick shift” while trying to navigate using such a cognitively overloaded screen and you’ll easily understand which Jobs To Be Done need to be prioritised, which can be good to have and which need to be canned completely. #productmanagement #gmaps #designthinking

112

21 Comments

A prominent Indian builder bought land in Ayodhya before the Ram Mandir announcement, hired Bollywood stars, and ran full-page ads, raising property prices by 900%. Now, buyers can't sell. This tarnishes India's real estate reputation, as people are often tricked into bad investments. However, real estate can be a good investment if you understand these points: 1. **Builder Control**: Builders sell high and buy low. Avoid buying overinflated properties for good returns. 2. **Clueless Buyers**: Ensure rental yield is above 4%. On a 1Cr property, rent should be at least 4 Lakhs annually. Overpaying leads to poor returns. 3. **Inflated Prices**: Selling an overinflated property is difficult. Buyers will investigate rental yields. Builders create FOMO to sell. You need experience and good judgment. Buy real estate only when it mathematically makes sense. CREDIT: Akshat Shrivastava #RealEstate #Investment #Property #India #RealEstateTips #BuyerBeware #PropertyMarket #InvestmentStrategy #Ayodhya #FOMO Rishiyank Sharma

7

2 Comments

Spent some time today going through Meesho's blog on how they reduced 90% of their Redis infra cost and it is filled with interesting details ⚡ The blog covers their design, key decisions, and nuances around picking the right data structure and tuning Redis the right way. Give it a read, it is pretty informative. The high-level overview of it 1. key-value based initial design leading to bloated memory 2. moving to Redis Hash 3. it still bloats memory, but beyond a threshold 4. internals details on how Redis uses ziplist to implement hash 5. tuning `hash-max-ziplist-entries` configuration 6. minimizing calls to Redis using Lua script I have seen many people adopting KV based approach and then throwing money on infra to handle scale. But Meesho's blog highlighted how knowing database tunables can help you significantly reduce your infra costs. give it a read - https://lnkd.in/gp9N3JMh To be honest, this is a pretty interesting blog for anyone using Redis in production. I made a similar blunder back in 2020 and then spent 2 weeks optimizing it, so a hard relate :) #AsliEngineering #SystemDesign

663

16 Comments

If you are attending ICPR 2024 in the next 5 days (Dec 01-05), then do visit our poster titled: "waLLMartCache: A Distributed, Multi-Tenant and Enhanced Semantic Caching System for LLMs”; my co-author Binay Gupta will be presenting it. Our system waLLMartCache advances the GPTCache framework by incorporating the following features: 1. We introduce the support for a new database Redis in GPTCache – this is used as L2 cache in our designed system. 2. Currently, GPTCache is implemented to be run on a single node which we enhance to span across multiple nodes to handle industry-scale requests and consequently, we also designed a distributed eviction manager. 3. We further create partitions for individual tenants (clients) so that these can be hosted together while maintaining semantic separations. 4. We present a decision engine (i.e., an enhanced semantic similarity checker) that decides whether to cache an LLM response based on retail business use-cases. 5. We showcase that loading FAQs (which can be set to be stored persistently in the memory) while booting the LLM cache can be a simple yet effective strategy to boost cache hits significantly. Some of these contributions have already been merged with the open-source GPTCache code. A big shout out to my other co-authors: Soumik Dasgupta, Anurag Wagh, Lalitdutt Parsai, Geet Vudata, Shally Sangal, Sohom Majumdar, Hema Rajesh, Anirban Chatterjee. #GenAI #LLMs #MachineLearning #ArtificialIntelligence #Industry #Research #Conference #Paper

114

3 Comments

📉 The Fall of BYJU'S: A Cautionary Tale for Indian Parents As Byju's valuation plummets from a staggering $22 billion to virtually zero (ref in the comments section), it becomes imperative to examine the practices that have led to such a downfall. Unfortunately, these are not just corporate missteps but issues that deeply affect individual lives, particularly those of unsuspecting parents in Tier 2 and Tier 3 towns. One such case involves my relative, who has been entangled in an ordeal with Byju's that perfectly encapsulates the company's current troubles. Despite raising a cancellation request in April'24, Byju's has continued to charge them for both April and May (June they could not deduct because my relative did not put enough money in it). The BYJU'S support staff, which should be a beacon of assistance, has been unresponsive and unhelpful. This negligence has forced my relative to consider closing their primary bank account, a drastic step that disrupts their everyday transactions and financial stability. As per their support staff - they can not assure whether this deduction will stop or not even when the cancellation request has been raised, hence they are left with no choice but to close their primary bank account. Even after closing the bank account, my relative is still being relentlessly chased by Byju's recovery department to pay the subscription fees/emi. The tone and persistence of these recovery agents resemble that of extortionists, putting immense mental pressure on my relative. This aggressive and harassing behavior is unacceptable and further underscores the company's operational failures. Support Ticket # - 5910087 Such practices are not just administrative lapses but a betrayal of trust. Byju's was once seen as a revolutionary force in education, promising quality learning to children across India. However, the company's inability to handle basic customer service issues speaks volumes about its operational failures. Parents, particularly in Tier 2 and Tier 3 cities, invest in educational services like Byju's with the hope of securing a better future for their children. It is disheartening to see these very families being taken advantage of and left in distress. This erosion of trust is perhaps one of the fundamental reasons behind Byju's precipitous decline in valuation. 🛑 Parents, please exercise caution when dealing with Byju's. Always read the fine print and ensure you understand the terms and conditions before committing to any service. Your vigilance is the first line of defense against such malpractices. #Byjus #CustomerService #Education #EthicsInBusiness #Trust #IndianParents

7

3 Comments

Layoffs in past have proved that An average engineer in a high priority team is more valuable than a top engineer in a low priority team. Most of the layoffs are not relevant to performance but have a direct correlation with the projects team is working on. It is equally important to pick the right team along with the company. #softwareengineering #layoffs #programming

1,073

24 Comments

Could The New York Times be the next surprise on Zepto? I was going through NYT's website when I noticed their aggressive pricing (Annual subscription rate of Rs.600 vs. Rs.3000/Mint and Rs.2660/ET). Going through the traditional distribution channel of tying up with local distributors wouldn't make sense for NYT-too much effort, too little return. But what if they leverage Zepto's hyper-local delivery? To begin with, how big is the Indian market for NYT? We can use LiveMint as a proxy. The only reliable data point available online is that it had 6L reach in print as per IRS 2019. So assuming that number is still the same(because the decline in print media is offset by their expansion in new cities), Rs. 15/copy gives an annual revenue of ~3000 crore or $0.4B. Now I'm not saying NYT will capture that entire pie, but there's space, and it has nothing to lose. With NYT aggressively discounting for Indian readers, and Zepto catering to a similar urban, affluent demographic, it feels like a good opportunity. Plus, Zepto could gain a loyal, recurring user base that overlaps perfectly with NYT’s target market. Zepto will have to figure out the printing part which could simply be tying up with some existing printing press. NYT could later even expand to larger deals with corporates/airports/hotels once it has set it's foot in the game. Scheduled delivery apps would be best for this, like country delight, but they haven't reached the scale yet. I'm assuming there must be a reason why Qcomm players have not entered selling even premium newspapers, which have relatively higher margin compared to a Rs.4 local newspaper. It could be low margins, or the fact that it only sells in the morning, or that it has an expiry date of 1 day. But I do feel its worth giving a shot because why not. What do you think?

17

5 Comments

Cofounder Conflicts 😞 ! Everyone owns everything, no one owns anything! I agree so much, Aman! Many teams suffer because everyone is doing everything and unfortunately even when everyone on the team wants to take ownership and be accountable, by design they are unable to. Clear divisions from Day 0 aren’t just a good practice—they're essential. I would go ahead and even say that don't even wait for traction. Do it early, align responsibilities, and set clear metrics for each function. Not only does this create accountability, but it also prevents the extended teams from feeling confused about who to take directions from. Sure, co-founders can contribute across the board (they should!), but there must always be a primary owner for every key function and metric. It keeps the vision sharp and execution smoother. If you are a founder or someone who worked in an early stage startup, do share if you witnessed this? How did you tackle it? Let’s discuss! 👇 #Startups #CoFounders #Accountability #Growth #Leadership #Entrepreneurship #India

20

2 Comments

Intuit leveraged K8sGPT and GenAI in their on-call process, resulting in improved issue detection, debugging, and resolution. 🚀 On-call engineers often face frequent alerts, tedious log scanning, and multiple team handoffs to resolve issues. Often, few teams have high operations load and engineers face fatigue resulting in low productivity. 😌 Intuit was facing similar issues with their complex Kubernetes deployment comprising 325+ clusters and 7000 services. They came up with a methodical approach to solve the problem through GenAI. What were the goals of the project ? 👉 Reduce the Mean Time To Detect (MTTD) an issue 👉 Reduce the Mean Time To Remediate (MTTR) an issue How did they solve the problem ? 1️⃣ Golden cluster signals - Single view of cluster's health and availability. It helped engineers to detect the bad clusters quickly reducing MTTD. 2️⃣ Debugging with K8s GPT - Adopted open-source K8s GPT to analyze root cause of failures. K8s GPT has integration with Open AI and other models. 3️⃣ Remediation with GenAI - Since public LLMs don't have Intuit specific data, they hosted local models and augmented them with Intuit specific data (RAG) that suggested remediation Intuit is planning to extend the tool and use it for other use cases, such as traffic and Java virtual machine debugging. They have published a detailed blog on how they adopted K8s GPT and the challenges they faced. You can find the link in the first comment. 👇 On-call weeks are challenging, but investing in GenAI can streamline the process and improve work-life balance. Share in the comments if you've used GenAI to improve your on-call process. Also, explore how K8sGPT could boost your team's operational efficiency. Also, folks who are interested to know more about K8s GPT can go to the github link, understand the code and start contributing. 🔥 #chatGPT #GenAI #softwareengineering

56

9 Comments

One Android phone, one whiteboard, and one marker = $2.8 billion. Alakh Pandey's PhysicsWallah just raised $210mn at a $2.8bn valuation - 2.5X higher than their last round, with revenues notching the $250mn mark. The crazier part: founders and management STILL own 85% of the company. From being unable to pay for school and selling his family home, to founding India's 101st unicorn, Alakh Pandey's story has been something else. Since he was a child, Alakh had always wanted to study engineering at IIT. His parents couldn't afford to send him - and they sold their house so that he could find his way at a local engineering college. To make ends meet and buy back his family home, Alakh started teaching on the side. He began tutoring children whenever he was off from school, and almost every single rupee earned went into savings. But there was one problem. The tutoring business wasn't growing fast enough. So in 2016, Alakh decided to quit his teaching kids and focus on YouTube full-time. He started teaching under the name PhysicsWallah, and things took off. He quickly amassed a following that loved his style of teaching - so Alakh decided to monetize. PhysicsWallah turned into a moonshot when he introduced a 3500Rs. (US$43) IIT Prep Package. The cheapest option out there, taught by an amazing tutor, for India's most competitive exam. What's not to love? Subscribers came trickling in, PhysicsWallah's bank account started filling up, and Alakh Pandey's name became a staple for every young Indian aiming to go to IIT. Ed-tech platform Unacademy saw this, and offered him an annual salary of $5m+ every year to teach on their platform. Alakh said no. PhysicsWallah set off on a hunt for investors, and the rest is history. It's staggering how a whiteboard, a marker, and a camera took a humble man from the verge of bankruptcy to one of India's biggest household names.

154

13 Comments

In 2024, 38% of India's top institute's students were unemployed, from all #IITs combined. One distrubing fact apart from low placement rate I found is students from #IITs, #NITs, #BITs at times reject offers which is just bit less than their other friends from same colleges, EGO 🥲 everyone wants CRORES CTC. I'm not saying join a company for 1 LPA package, but please understand, when global situation is bad, you don't get 1 Crore CTC for 100 students, don't get biased by seeing YouTube videos 🙄😒 The value a fresher can add to a company is not that much considering capabilities, extremely changing technology, lower investments and wars going on, that's a fact, either understand it or suffer due to EGO. Maybe you can join some company with 5-10 LPA as fresher and PERFORM WELL for 1-3 years and than SWITCH based on skills, based that companies do invest on you. Please remember those 1 Crore fresher salaries or even 30, 40, 50 LPA CTCs come with 3-4 years stocks option, so, companies think long term and pay you based on that. And please remember, all the FAANGM companies were start ups once upon a time 🤏 Maybe as a fresher, join some small start up and help them become bigger, your salary will also grow exponentially 🙂🔥💯💸💸💸 I know getting into top colleges takes extreme hard work, getting placement again takes huge hard work but ..... sometimes adjusting according to situation is a smarter option. Let me tell you a fact, I currently have 3 Years and 10 months experience as a backend developer and this is my 3rd company, I've switched companies and worked in multiple products. Based on this, I can assure you that after a certain salary, more and more hike just becomes a number, now alongwith salary, I give equal importance to my career growth, just joining a brand doesn't matter to me. My lifestyle is controlled by me, not my CTC, is has dependency but not 100% 🙂 I've rejected multiple product based companies offers to join a smaller company, my second company because of better work expectations. I left that due to my work from office preference within few months, I still love WORKING FROM OFFICE. The reason of telling this personal information in this post is not to show off my salary, it is to tell my juniors and freshers, please understand "MONEY IS NOT EVERYTHING". Join startups and help them grow, you won't be unemployed for longer or underpaid based on SKILLS 🔥💸 #placements #freshers #iit #nit #bit https://lnkd.in/gk6WV5bW

4

📌 𝐇𝐢𝐠𝐡 𝐩𝐚𝐜𝐤𝐚𝐠𝐞 𝐬𝐚𝐥𝐚𝐫𝐲 - wants to mint money ? 📌𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈 / 𝐋𝐋𝐌 - want to learn Gen AI? 📌 𝐓𝐨𝐩 𝐓𝐞𝐜𝐡 𝐜𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 / 𝐅𝐀𝐀𝐍𝐆- want to land up in your dream company ? 👉You can do it all in just 1 click and just at the cost of golgappa (with Diwali discount 🪔). I will send you Discount code on DM if you comment “Happy Diwali” and / “interested” 🧑‍🔬If you are a Data Science aspirant and looking for job roles like Machine Leaning Engineers, Al Scientist, Data Scientist, Generative Al roles, etc or are willing to make acareer switch or want to sky-rocket your career, I can assist you on ✅Resume building ✅Career Guidance ✅Mock interview ✅Acquiring a Letter of Recommendation ✅Interview preparation tips ✅How to convert into Data science job roles from Data Analytics or from other non-IT/IT-related jobs 💁‍♂️And the best way to connect and assist you is 1:1s topmate.io https://lnkd.in/gijrQDK3 🧐Don't wander and waste your time at 100 different things and resources. #topmate #ai #generativeai #llm #diwali #discount

34

1 Comment