Ayan Sen

MCP - a open protocol for AI (LLM) to talk to other tools / systems / applications. Naive, but don’t judge me, I was judging AI capabilities with me being Naive   - So I asked Claude AI to create an MCP - An MCP the AI itself will use - Gave a website address to which MCP had to be created (with public APIs) - It Generated extensive, detailed code - I dwelled on it, fine tuned for my preference / requirements, it went ahead and did a bit more than I asked for. Website was analysed in minutes (amusing and scary) - I asked it to package it as executable package - I had configured file system MCP - So, Claude leveraged that and created a python project in filesystem and created Project with proper files and folder structure - Here comes Claude code (CC) - This python project of MCP server, was not in a running state - Asked CC to make the code runnable (of course I was in control (manual loop) of the changes it was making) - MCP server to run in local system (MacOs) needed some changes in trivial python code, CC made the changes - Interesting side part, during this whole process, it was also creating documentation, to install server, to run server and the MCP config for other tools - At times I had to say update the documentation - OOPS !! Code got changed, but what was changed ? - I asked CC to make this project version controlled. Oh my Devaa (God), it converted the project into a Git repo (local) - Created git related files, Auto committed all files. All I had to do was, say yes, go ahead - Well, we got side tracked, lets get back to MCP - So, now I asked CC to run the server (lazy to copy the command and run it in next tab terminal) - Obviously.. It had issues, and obviously it auto corrected the code - It wrote test cases, wrote test start script, tested it, committed the code and then - now it tells me, 'All fixed, Want to run?', Howdhu maga howdhu, run (yes dude yes, run) - MCP server started. I updated MCP configuration and restarted Claude Desktop - pssssttt….. curios? - :( Error: MCP server couldn’t be started - Again me being extremely (purposefully) lazy, told CC, you are being dumb, fix the startup error - I guess, I hurt its Ego. - It asked permissions for MCP server log file. whaaaat really ?, it wants to read log files now, wow. - It test ran with its test scripts, read the log, analysed like a nerd, updated the test files and code, reran until it was fixed and updated the documentation and also committed the code - What can I say, I was watching Christopher Nolan movie level events. - Finally, MCP worked, I tested few things, results were as expected.   It ain’t over, my architectural / Engineering hormones kicked in. How about containerisation ? How about productionising it ? How about Security & Authentication ? How about Performance & Scalability ? How about Operational Readiness ? How about AI Security, Gaurdrails ? Stay tuned, I will be back 🙂

14

Salim FEKU not able to digest NoSQL databases are converging to PostgreSQL. Setting: A chai ki dukaan (tea stall) in Hyderabad. Salim Feku is holding forth, gesturing wildly with his chai glass. (Salim Feku sips his chai, then puts the glass down with a flourish, looking around at an imagined audience.) SALIM FEKU: Arre bhai, aaj kal ki duniya mein kya chal ra? Sab bolte, "PostgreSQL, PostgreSQL!" Kya hai re usmein? Apun bolta, woh to puraane zamane ka ghoda hai! (He leans in conspirantly, lowering his voice slightly but still loud enough for everyone to hear.) SALIM FEKU: Suno tum meri baat, agar tumko asli data rakhna hai, jo kabhi nikalta nai, jo tezz bhaagta, toh tumko NoSQL hona bhai! Yeh PostgreSQL, woh to apna aadmi, kaisa... sara din join lagate baithta. Ek table se dusra table, dusre se teesra... Arre baba, jab tak woh join lagake data nikalta, apun chai pi ke doosra dhanda kar leta! (He chuckles, shaking his head.) SALIM FEKU: Apun ke paas kya hai? MongoDB! Arre kya cheez hai woh! Tum feko usmein kuch bhi! Document feko, JSON feko, apun ka data, seedha andar. Koi tension nai. Woh PostgreSQL, usko pehle batao 'column' kya hai, 'row' kya hai, 'schema' kya hai. Arre baba, itna sochne mein hi system slow ho jata! (He snaps his fingers.) SALIM FEKU: Aur suno, jab apun ka traffic aata, haan? Croreon log aate, kya karte? Woh PostgreSQL to ek machine pe baitha rehta. Apun ka Cassandra, DynamoDB... yeh log kya karte? "Arre bhai, thoda data tu le le, thoda data main le leta," bolke baant lete. Shaana log! Sab milke kaam karte, isliye apna system kabhi rukta nai, bhai! Woh PostgreSQL ko to pasina aa jata! (He gestures expansively, like he's spreading data everywhere.) SALIM FEKU: Woh bolte, "Hum consistent hai!" Arre haan, theek hai. Lekin consistency ke chakkar mein kya, apna customer wait karte baithta! Apun ka NoSQL? Woh bolta, "Thoda late aaya to kya hua? Aayega to sahi! Customer khush hona, jaldi result aana!" Zara sa "eventually consistent" hua to kya hua? Duniya to chalte rehti na! (He lowers his voice again, for the grand finale.) SALIM FEKU: Toh bhai, agar tumko naya, tezz, bina tension ka system banana hai... toh woh puraane dimaag ke relational ko choro! Apun ka NoSQL, woh hai asli king! Baaki sab to... chillar hain! (Salim Feku takes another proud sip of his chai, looking very pleased with his pronouncement.Just then, his eyes fall on someone approaching the chai stall. His confident smirk vanishes, replaced by a look of surprise, then forced cordiality.) SALIM FEKU: (Suddenly a soft, almost meek tone) Arre! Kaun? Ismail Bhai... PostgreSQL wale!?! (He rushes forward, abandoning his lecture, ) SALIM FEKU: Aao Miya! Kaisa hai tum? Tumhari hi raah dekh raha tha main! Kaisa, sab khairiyat? Woh, woh main to... ...main to bas logon ko bata raha tha ki purani cheezo ki reliability kitni zaroori hai! Haan! Tumhari hi baat kar raha tha, bhai! Maine suna..tum bhi distributed ho gaye!!!

8

Building an agent or RAG system with high accuracy and low cost and latency can be a daunting problem. Between the choice of language models, prompts, depth of reasoning, text splitters, retrievers etc., the options to explore can be innumerous. Hence we build Syftr at DataRobot. Syftr auto-magically searches over the large set of hyperparameters to find the pareto-optimal frontier of configurations given accuracy, cost and latency constraints based on the needs of the agent. Syftr uses Bayesian optimization to efficiently search over the hyperparameter combinations. The set of combinations can be >> 10^23, you might as well give up optimizing this manually. Syftr uses a novel pruning technique that stops early with non-promising search paths. This is how it can efficiently identify the optimal frontier. The best part is syftr is fully open-source. Clone the repo and try it out yourself. It can be configured to support various LLMs, prompt strategies, RAG configs etc. Check it out yourself here: https://lnkd.in/gbPnNW_D DataRobot was a pioneer in AutoML with predictive models, that DNA seeded the idea to build Syftr at DataRobot. Special shoutout Debadeepta Dey for leading Syftr and giving life to this idea. Looking forward to get Syftr natively in DataRobot.

78

1 Comment

Stop Asking “Which LLM Is Bigger?” Ask: “Which LLM Understands Your Customer?” India has 22+ official languages, countless dialects, and millions of daily code-mixed interactions. Yet many enterprises deploy global LLMs trained largely on English-heavy datasets — expecting perfect customer intelligence. That’s the strategy gap. Global LLM strengths: ✔ Broad knowledge ✔ Strong English benchmarks India-facing challenges: – Sarcasm misclassification in Hinglish – Accent-driven ASR inaccuracies – Loss of regional nuance – Compliance & data residency complexity India-trained LLMs — such as those built by Sarvam AI — are optimized for: ✔ Multilingual Indian corpora ✔ Code-mixed language patterns ✔ Regional accent tuning ✔ Indian script OCR ✔ Context-aware sentiment & sarcasm detection This isn’t nationalism. It’s optimization. AI selection is no longer about model size. It’s about contextual intelligence, regulatory alignment, and ROI precision. In India, competitive advantage won’t come from the largest model — but from the most context-aware one.

9

If agentic systems can generate code (or content, workflows, configs) at near-zero marginal cost, then “more output” cannot be the north star. The stock(https://lnkd.in/gXYbUxTr) will explode. The real design question becomes: how do we build agentic systems where stock health — not inflow velocity — is the primary objective? Here’s a simple framework. First, separate Flow Optimization from Stock Optimization. Most agents today are optimized for flow: number of PRs, tasks closed, tokens generated, features shipped. A stock-health system explicitly measures the condition of the accumulated asset — cohesion, complexity, defect density, infra cost, latency, security exposure, change failure rate. If it’s code, track architectural entropy and long-term maintainability. If it’s data, track lineage integrity and semantic consistency. What gets measured becomes the system’s gravity. Second, embed a Reflexive Feedback Loop. Every act of generation must create a shadow of evaluation. Agents shouldn’t just write — they should simulate impact: dependency ripple, test coverage delta, performance implications, cost drift. In systems terms, generation must automatically trigger balancing signals. If complexity crosses a threshold, refactoring tasks are auto-created. If coupling increases, modularization is incentivized. The agent is rewarded not just for adding, but for simplifying. Third, align incentives around Net Stock Quality, not Gross Output. Reward deletion. Reward consolidation. Reward reduction in cognitive load. An agent that removes 1,000 lines of unnecessary code should score higher than one that adds 10,000. This flips the reinforcing loop. Instead of “more code → more work → more code,” you get “more complexity → more pruning → restored equilibrium.” Finally, introduce a Steward Layer — human or meta-agent — whose sole mandate is long-horizon coherence. Not velocity. Not tickets closed. But architectural narrative. This layer ensures that local optimizations by agents don’t fragment the system over time. Think of it as system memory and taste. In short: build agents that generate, evaluate, prune, and remember. When stock health becomes the objective function, reinforcement and balancing loops coexist intentionally — and abundance becomes leverage instead of entropy. #systemsthinking #ai #agenticai #systems

16

System Design Part 2 Continuing my System Design Blog Series, this post is a deep dive into the Low-Level Design of WhatsApp. 📱 From handling 2 billion daily messages to ensuring high availability, we’ll walk step by step through how WhatsApp is architected at scale. 🚀 High-level estimate Here is our high-level estimate: TypeEstimateDaily active users: (DAU)50 million Requests per second (RPS) : 24K/s Storage (per day) : ~10.2 TB Storage (10 years) : ~38 PB Bandwidth : ~120 MB/s

6

About a month ago, I had the opportunity to speak at Redis Delhi, alongside some eminent industry leaders — Suman Guha (Tata CLiQ / Tata Digital), Murali Mohan Chakravarthy (Entain), and Rhythm Goyal (Monotype) My session ran for ~25 minutes and focused on how Redis has evolved far beyond its original role as an in-memory cache, and how teams are using it today to solve modern, large-scale problems. We discussed: - Vector search and its role in modern search systems - Bloom filters for efficiently checking user existence at scale (probability at scale!) - Sorted Sets as a practical primitive for rate limiting - Why in-memory systems like Redis became popular for agentic workflows instead of disk storage, aka GPU runtime costs What stood out the most was the audience—an enthusiastic, well-curated mix of experienced engineers and young, fast-growing talent. The questions and discussions made it a genuinely engaging session. It’s always a pleasure learning and collaborating with Jatin Gupta, Omkar Konnur, Justine Hermoso, and Rakhilaa Rajasheker. Thanks to the Redis Delhi team for having me and everyone who joined 🙌

55

4 Comments

📯With 𝗦𝗽𝗿𝗶𝗻𝗴 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝟳.𝟬 set to release by end of 2025, one of the major excepted feature is improved native image support for 𝗚𝗿𝗮𝗮𝗹𝗩𝗠. While Spring started supporting native images from version 6.x via Spring AOT hints, version 7.0 pushes this further by deepening support for:  • Spring AOT contributions, making native image builds more robust and reliable.  • Removing deprecated methods and aligning more cleanly with GraalVM expectations. The key to choosing 𝗚𝗿𝗮𝗮𝗹𝗩𝗠 𝗼𝘃𝗲𝗿 𝗢𝗽𝗲𝗻𝗝𝗗𝗞 however depends on understanding 𝘄𝗼𝗿𝗸𝗹𝗼𝗮𝗱, 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲 𝗮𝗻𝗱 𝘁𝗿𝗮𝗱𝗲-𝗼𝗳𝗳𝘀 - which I have documented in this post to best of my knowledge. ⚖️ Key Trade-offs ✅ Choose GraalVM Native if:  • Cold start speed is business-critical (serverless, on-demand scaling).  • Memory footprint must be minimized (edge devices, containers with hard limits).  • You’re building batch/CLI tools that should run instantly.  • You can tolerate longer build times in CI/CD. ✅ Choose JVM (HotSpot/OpenJDK) if:  • App runs 24/7 and throughput > startup time (e.g., APIs, Kafka, banking apps).  • Heavy runtime reflection, proxies, or dynamic class loading.  • Developer productivity (fast builds, hot reload, live debugging) is a priority.  • You need mature monitoring, profiling, and ops tooling. Hope this helps! Which are the use cases where you think GraalVM would prove better over OpenJDK? #OpenJDK #GraalVM #AOT #SpringFramework #Java #SoftwareDevelopment #JavaEngineering

5

Ever had a Spark job crash with OutOfMemory errors — even when you allocated “enough” memory? Spark's memory model can be confusing. Here’s a practical example to help you make sense of it. You submit a job with the following Spark config: Ø spark-submit \ Ø --deploy-mode cluster \ Ø --master yarn \ Ø --num-executors 1 \ Ø --executor-cores 4 \ Ø --executor-memory 2G \ Ø --conf spark.dynamicAlllocation.enabled=false Here's how Spark actually allocates that memory: Total Executor Memory = 2GB 🔸 300MB → Reserved for Spark engine 🔸 ~1.7GB left for use 🔹 60% → Unified memory (execution +storage) ≈ 1GB 🔹 40% → User memory ≈ 700MB + Overhead Memory (outside JVM) max(10%of executor memory, 384MB) → Added on top of 2GB The boundary between execution memory and storage memory in Spark is not strict. Spark uses a unified memory model, meaning execution and storage memory share the same pool. Based on current needs and availability, execution memory can borrow from storage and vice versa. This flexible model helps optimize memory usage, but it's important to note the eviction policy: If execution memory has already occupied a large portion (say,70%), and storage later needs more space, storage cannot evict execution memory. However, execution can evict storage blocks (like cached data) to make room for its operations. ------------------------------------------ Memory Types Overhead Memory - Memory allocated for VM related overhead. Reserved Memory - Memory allocated for Spark Engine. Storage Memory - Memory for Cache and Persist operations. Execution Memory - Temporary memory required for execution of Join | Sort | Shuffle | Aggregations ... User Memory - Memory for RDD related operations, and user defined data-structures. -------------------------------- Configuration properties used to set the values for following categories of memory overhead - spark.executor.memoryOverhead heap - spark.executor.memory offHeap - spark.executor.offHeap.size pyspark - spark.executor.pyspark.memory #ApacheSpark #DataEngineering #SparkOptimization #BigData #MemoryManagement #PySpark #CloudComputing #DataEngineer #datarngineer

16

Google Antigravity exposes the critical flaw in autonomous AI agent architectures: trust based governance at scale. Antigravity represents a major shift in how developers work. Instead of writing code line by line, you delegate entire tasks to autonomous agents that can modify files, run tests, browse the web, and execute changes across your codebase in parallel. The problem? These agents operate on trust, not proof. One developer reported: On Day 3, an agent confidently refactored a utility function and silently deleted a critical edge case check. This is not a bug. This is the inevitable result of autonomous agents operating without cryptographic authority validation. When you have three agents working asynchronously across different files, two critical questions emerge: how do you enforce what each agent is authorized to do, and when multiple agents coordinate on shared resources, how do you maintain isolation between their operations? Policy based guardrails do not work at this scale. I have solved both problems through complementary cryptographic architectures. For individual agent authorization, my 13 layer cryptographic governance system validates AI agent authority mathematically before execution. For multi agent coordination, my YIN COLLAB architecture implements Agent Specific Compliance Tokens with per agent privacy isolation maintaining cryptographic boundaries preventing cross contamination. Every action carries immutable proof of authorization. Every decision boundary is pre validated cryptographically. 26 USPTO patents. 2,330 claims. Validated with 640x timing resistance and 500 plus concurrent agent support with sub 15ms latency. Mathematical proof, not policy promises. For developers using Antigravity, Cursor, or any autonomous agent platform: the question is whether you can prove mathematically that unauthorized operations are impossible and that multi agent coordination maintains isolation. Because as agent orchestration becomes the dominant development paradigm, the liability surface expands exponentially. One misconfigured agent with access to production systems is an organizational failure. One agent leaking sensitive data to another agent through shared context is an architectural vulnerability. Making it mathematically impossible for agents to violate boundaries is the only governance model that scales. Autonomous agents are the future of software development. Cryptographic governance is the only way to make that future safe. https://lnkd.in/ejPREk9D #GoogleAntigravity #AIGovernance #AutonomousAgents #Cybersecurity #AIAgents #DeveloperTools #CryptographicSecurity #ZeroTrust #AICompliance #SoftwareDevelopment #TechInnovation #AIEthics

2