About
Site reliability engineer for some of the largest scale production environments (on-prem…
Activity
2K followers
-
Sandip Bhattacharya shared thisI am afraid of the day when someone runs this analysis on LinkedIn 😛 https://lnkd.in/eCan26Kp
-
Sandip Bhattacharya shared thisReally like this model. This is now my goto model (via fireworks of course) for all non-subscription AI usage.Sandip Bhattacharya shared thisKimi K2.6 is live on Fireworks as a Day-0 launch partner! K2.5 was the base for standout models like Cursor’s Composer 2 and was the most popular model on our training platform. K2.6 on Fireworks raises the bar again. → Optimized across the stack, from custom speculators to heterogeneous hardware support across NVIDIA and AMD. → Day-0 serverless support is live, and coming soon to Fire Pass (stay tuned). → Unlock new use cases with capabilities for 12+ hour autonomous runs and 4,000+ tool calls. → $0.95 input / $4.00 output per 1M tok Get started today → https://lnkd.in/gqXXpZrg
-
Sandip Bhattacharya shared thisPosition open in Canada, USA, UK. Ceph is the critical filesystem layer in all our Kubernetes environments, and we have a lot of really interesting problems to solve at scale!Sandip Bhattacharya shared thisWe’re hiring a Senior Ceph Engineer at Index Exchange! 🚀 You’ll be working on distributed storage systems powering a platform that processes 550B+ real-time auctions daily and 2PB of data every day, partnering closely with an excellent leader, Sahil Yusuf. We're looking for someone who: • Has experience with Ceph • Has worked in high-scale, production environments • Enjoys solving reliability and performance challenges Apply here: https://lnkd.in/eHXeMzYp
-
Sandip Bhattacharya shared thisThis sounds urgent. 🏃♂️OpenAI Issues Urgent Warning: macOS Users Must Update ChatGPT & Codex ImmediatelyOpenAI Issues Urgent Warning: macOS Users Must Update ChatGPT & Codex Immediately
-
Sandip Bhattacharya shared thisOne thing I like about #ClaudeCode is that it has no opinion about what you can use it for. I use it for coding, sysadmin work, writing obsidian notes, everything. #OpenCode, my second best agentic cli, on the other hand has annoying opinions about what you should use it for - "I am a software engineer only, thanks."
-
Sandip Bhattacharya posted thisIt is increasingly hard not to see the parallels between how the core behaviour of LLMs are not entirely explainable by their very own creators (why do emergent behaviour happen at scale), and the emerging trend of software being built with LLM generated code that is increasingly unsupervised.
-
Sandip Bhattacharya shared thisI have heard about this trend of _no IDE_ and I still don't entirely get it. So is the expectation that either an engineer only looks at diffs in code, or not do code reviews at all? IDEs give better context to change sets even in an AI driven workflow, imo. I frequently spot inconsistencies in a different part of the code while reviewing changes from the agent in the IDE window. Sometimes docs are out of sync, sometimes neighbouring code turns redundant. IDEs are still important for humans managing code. We can handle larger context, while the agents are still desperately trying to grep through code to fit into their tiny context windows. Humans are imperfect, and have less mechanical capabilities than agents, but they are better stewards of the code. It is hard keeping up with rapidly changing coding practices already, but I am deeply uncomfortable with this wish of turning engineers into agent managers. There is a whole bunch of articles emerging about skill atrophy in engineers who spent less time being involved with actual code.Sandip Bhattacharya shared thisThis week I told my engineering org two seemingly insane things: 1. Delete your IDE. 2. Write 0 lines of code. There was immediate skepticism. A little “wtf, he's crazy.” Fair. Then something interesting happened: people leaned in. It was obvious there was no faster way to learn. Why? If you want an organization to operate AI- and agent-first, you can’t layer new tools on top of an old job definition. You have to change the work itself. IDEs are comfort blankets. Writing code is muscle memory. Both anchor you to a job that no longer exists. The new job is direct agents and shape systems to ship outcomes. We’ve been aggressively using AI since late 2024—long before it was fashionable—and still at times feel behind. The issue was never tooling. It was mindset. Maybe the real bottleneck has been us, humans, the whole time. The invariant: build things users care about. Everything else is just how you get there. If you want to change the game, you have to actually play it.
-
Sandip Bhattacharya reposted thisSandip Bhattacharya reposted thisThat feeling after migrating the control plane nodes on a 𝟱𝟬 𝗣𝗲𝘁𝗮𝗯𝘆𝘁𝗲 𝗰𝗹𝘂𝘀𝘁𝗲𝗿 while all data pipelines were live, 𝟳 𝗺𝗶𝗹𝗹𝗶𝗼𝗻 𝗲𝘃𝗲𝗻𝘁𝘀/𝘀𝗲𝗰 flowing through, and not a single user impact. All of this happened at 7:30 PM on a Friday just 30 minutes before the Jays & Dodgers World Series game we were all excited about. Yet the focus, ownership, and teamwork never dipped. 🚀 . Thanks to Everyone for clean execution. 𝗘𝗻𝗷𝗼𝘆 𝘆𝗼𝘂𝗿 𝘄𝗼𝗿𝗸 𝗮𝗻𝗱 𝗔𝗽𝗽𝗿𝗲𝗰𝗶𝗮𝘁𝗲 𝘆𝗼𝘂𝗿 𝘁𝗲𝗮𝗺 !! #teamwork #dataengineering
-
Sandip Bhattacharya shared thisLove this take.Sandip Bhattacharya shared thisHere's something remarkable: You almost *never* hear the most popular view of AI that the majority of people who work in tech have about technologies like LLMs. By popular request, I've tried to sum up that POV as succinctly as possible, and explain why many are afraid to tell a simpler, less dramatic story that contradicts the narrative that the billionaire tycoons are trying to force on the world. https://lnkd.in/e5pjGvVP
Sandip Bhattacharya commented on a post
21h
Now if all the people could only agree as to where in the filesystem to keep your worktrees.
Sandip Bhattacharya commented on a post
3d
When is GitHub copilot adding more model providers? Isn’t it a win-win for both sides when we can use pretty good qwen/kimi/gemma models right from within copilot instead of byok and getting more bang for the buck. They are an order of magnitude cheaper. And you have US providers like fireworks etc if provenance is an issue
-
Sandip Bhattacharya liked thisSandip Bhattacharya liked thisApril was a pretty strong month for open-weight LLM architecture releases: 1. Gemma 4 Continues the local/global attention recipe with sliding window attention, which is a classic yet "easy" way to extend context while making it cheaper than full attention. 2. GLM-5.1 A large MoE model that based on DeepSeek V3.2 with multi-head latent attention (MLA) and DeepSeek Sparse Attention 3. Qwen3.6 Keeps the Qwen3.5-style hybrid stack. I.e., mostly Gated DeltaNet / linear-attention-style layers, with periodic full attention layers. Next to the 35B-A3B MoE variant there now also is a 27B dense variant. 4. Kimi K2.6 A 1T-scale sparse MoE model with DeepSeek-style MLA and only a small fraction of parameters active per token. It is another good example of how frontier-scale models keep improving without major architecture changes (coming from DeepSeek V3) except increasing the size. (Lots of performance improvements are due to the training recipe.) 5. DeepSeek V4 Probably the most architecture-heavy release of the month. The Pro and Flash variants add manifold-constrained hyper-connections, compressed attention variants, hash-based routing in the early layers, and a 1M-token context window. All of these are now added to the LLM Architecture Gallery. I am keeping the notes short for now, especially for DeepSeek V4, because the details around mHC and compressed attention deserve a more careful writeup. More once I am fully back in May.
-
Sandip Bhattacharya liked thisSandip Bhattacharya liked thisKimi K2.6 is live on Fireworks as a Day-0 launch partner! K2.5 was the base for standout models like Cursor’s Composer 2 and was the most popular model on our training platform. K2.6 on Fireworks raises the bar again. → Optimized across the stack, from custom speculators to heterogeneous hardware support across NVIDIA and AMD. → Day-0 serverless support is live, and coming soon to Fire Pass (stay tuned). → Unlock new use cases with capabilities for 12+ hour autonomous runs and 4,000+ tool calls. → $0.95 input / $4.00 output per 1M tok Get started today → https://lnkd.in/gqXXpZrg
-
Sandip Bhattacharya liked thisSandip Bhattacharya liked thisParadigm 2026 – Celebrating Legacy, Strengthening Connections On April 18, 2026, the Alumni Relations Committee, MBA—IISWBM, in collaboration with the Events Committee - IISWBM, proudly hosted PARADIGM 2026 at The HHI, Kolkata, bringing together esteemed alumni across batches for an evening of meaningful engagement and shared legacy. The event was graced by the esteemed presence of IISWBM’s Director, Prof. (Dr.) K. M. Agrawal, and Head of the MBA (Day & Evening) programme, Prof. (Dr.) Tanima Ray, whose guidance and encouragement continue to shape the institution’s academic excellence. We were also privileged to have a large number of respected faculty members in attendance, whose presence added immense value and warmth to the occasion. We were deeply honoured by the gracious presence of our distinguished alumni, whose participation made the evening truly enriching. Their insights, experiences and continued association with the institute remain a constant source of inspiration for our students and the IISWBM community as a whole. The evening fostered meaningful conversations, mentorship and a renewed sense of belonging—further strengthening the vibrant IISWBM alumni network. To our alumni and faculty—thank you for making PARADIGM 2026 truly memorable. #PARADIGM2026 #IISWBM #AlumniRelations #MBA
-
Sandip Bhattacharya liked thisSandip Bhattacharya liked this𝗧𝗵𝗲 𝗔𝗹𝘂𝗺𝗻𝗶 𝗥𝗲𝗹𝗮𝘁𝗶𝗼𝗻𝘀 𝗖𝗼𝗺𝗺𝗶𝘁𝘁𝗲𝗲 𝗼𝗳 𝗜𝗜𝗦𝗪𝗕𝗠-𝗠𝗕𝗔(𝗗𝗮𝘆) is glad to present the team representing the 𝗕𝗮𝘁𝗰𝗵 𝟮𝟬𝟮𝟰-𝟮𝟬𝟮𝟲 & 𝗕𝗮𝘁𝗰𝗵 𝟮𝟬𝟮𝟱-𝟮𝟬𝟮𝟳. Building on the responsibilities entrusted to us, we aspire to strengthen meaningful ties between the Institute and its alumni, while supporting initiatives that uphold the IISWBM spirit. As a new chapter begins, this team looks forward to contributing to a legacy of connection, collaboration, and community. 𝗪𝗶𝘁𝗵𝗶𝗻 𝗳𝗿𝗮𝗺𝗲:– ~ front row ~ Swarnima Chakraborty – 𝐇𝐞𝐚𝐝 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 Debasmita Das – 𝐂𝐫𝐞𝐚𝐭𝐢𝐯𝐞 𝐇𝐞𝐚𝐝 & 𝐓𝐫𝐞𝐚𝐬𝐮𝐫𝐞𝐫 Debarati Roy – 𝐏𝐮𝐛𝐥𝐢𝐜 𝐑𝐞𝐥𝐚𝐭𝐢𝐨𝐧𝐬 𝐇𝐞𝐚𝐝 & 𝐄𝐝𝐢𝐭𝐨𝐫 Spandan Bose – 𝐀𝐥𝐮𝐦𝐧𝐢 𝐄𝐯𝐞𝐧𝐭𝐬 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 ~ back row ~ Anushka Bhowmick – 𝐉𝐮𝐧𝐢𝐨𝐫 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 Ritabrata Mukherjee – 𝐉𝐮𝐧𝐢𝐨𝐫 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 Shirsha Gooptu – 𝐉𝐮𝐧𝐢𝐨𝐫 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 Sudeshna Karmakar – 𝐉𝐮𝐧𝐢𝐨𝐫 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 Here’s to a new year of connections, collaborations, and community.
Experience & Education
-
Index Exchange
View Sandip’s full experience
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Publications
-
Professional Apache Tomcat
Wrox
The first book on Apache Tomcat. Includes how to setup, configure and use.
Other authors -
Recommendations received
3 people have recommended Sandip
Join now to viewOther similar profiles
Explore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More