Notes on AI, ML, data & the engineering behind them.

Designing for Scale · LLM-as-a-Judge and Release Gates

How to ship generative AI to production without shipping a liability. Breaking subjective prompts into deterministic grading rubrics, automated evaluation pipelines, and multimodal scoring.

Designing for Scale · End-to-End Design Review I

Pulling the components together. How to trace a requirement from population estimates through capacity math to a defensible caching strategy, using a live design review as the frame.

Designing for Scale · End-to-End Design Review II

A second live design review, tearing apart an event ticketing system. Handling massive, instantaneous write spikes, distributed transactions, and inventory locks without destroying the database.

Designing for Scale · The Storage Decision Tree

Databases are not religions. Choosing between relational, wide-column, document, and blob storage by mapping the access pattern before looking at the engine.

Designing for Scale · CAP and the Consistency Spectrum

Why strong consistency is a latency penalty you choose to pay, why eventual consistency is not a defect, and how to use the CAP theorem to end distributed systems arguments.

Designing for Scale · URL Shortener

The classic system design starting point. It looks like a toy problem until you have to guarantee collision-free generation at scale while keeping read latency under ten milliseconds.

Designing for Scale · Distributed Rate Limiter

How to stop abuse without stopping legitimate traffic. An engineering breakdown of Token Bucket, Leaky Bucket, Sliding Window algorithms, and how to execute them across a fleet of servers without locking Redis.

Designing for Scale · Chat and Real-time Communication

Why HTTP fails for real-time delivery, how WebSockets change the load balancer math, and the architecture required to deliver a message to a million users simultaneously.

Designing for Scale · Newsfeed and Fan-out

Why reading from a database to render a feed is too slow, and how the fan-out-on-write model precomputes millions of feeds in memory before users even ask for them.

Designing for Scale · Typeahead and Search Autocomplete

How to return query suggestions in under fifty milliseconds while the user is still typing, using Tries, offline aggregations, and edge caching.

Designing for Scale · Notification Systems

Why sending a push notification is not a fire-and-forget API call. Handling rate limits, provider outages, deduplication, and the retry queues required to make delivery reliable.

Designing for Scale · Real-Time Analytics

How to count billions of events in real-time without crushing your database, using stream processing, time-window aggregations, and Lambda architecture.

Designing for Scale · Change Data Capture (CDC)

Why dual-writes fail, how the transaction log is the only true source of state, and the architecture required to stream database changes to search indexes and caches without losing data.

Designing for Scale · The Data Lakehouse

Why the data warehouse and data lake converged. Moving from expensive, proprietary compute-storage monoliths to open table formats like Iceberg, Hudi, and Delta Lake.

Designing for Scale · Metering and Billing

Why billing systems cannot drop a single event, the difference between at-least-once and exactly-once processing, and how to build idempotent pipelines that survive crashes without double-charging users.

Designing for Scale · The Feature Store

Bridging the gap between data engineering and machine learning. How to serve features for model training offline, and serve those exact same features for inference in five milliseconds online.

Designing for Scale · Model Serving

Why wrapping a PyTorch model in a Flask API is a prototype, not a production system. Handling GPU saturation, dynamic batching, and the difference between CPU and GPU scaling.

Designing for Scale · RAG and Vector Search

Why LLMs hallucinate, how Retrieval-Augmented Generation grounds them in reality, and the architecture required to execute semantic search over millions of documents in milliseconds.

Designing for Scale · Requirements to Architecture

From a one-line prompt to a defensible architecture: separating functional from non-functional requirements, quantifying who/what/how-many, and refusing to draw a single box until the constraints are on the board.

Designing for Scale · Estimation That Constrains

Back-of-envelope arithmetic is only worth doing when a number rules something out. How to derive requests per second, storage growth, bandwidth and working set — and how to tell a constraining estimate from a decorative one.

Kimi K3 From Zero to Deep — Math, Architecture, Training, and Systems

A zero-to-deep Kimi K3 course: prerequisite math and terminology, Transformers and MoE, KDA, AttnRes, vision, RL, training systems, inference, benchmarks, and a verified YouTube learning path.

Introduction to TensorFlow on Google Cloud

Diving deep into a Google Skill boost

Taking the Azure Fabric Ignite Edition Challenges to Complete

Microsoft Learn Challenge conducting a challenge to get good in few of the challenges which are super useful to complete to gain knowledge on Microsoft Fabric.

Exploring different services in GCP

Exploration and documentation of different services offered in GCP

Starting a company in India

Documenting the process of starting a company in india

Google 5 Day Gen AI course with interactive hands-on practice

Google and Kaggle provided good summary course on Gen AI , the blog contains details and highlights of the course.

AI Voice chatting to help with Customer support use-cases

Using current speech augmented LLMs (SpeechLLMs) with realtime voice modality to understand user issues and to provide support and solutions.

Audio to Video Generation Using Replit AI and Deploy as an Azure Webapp

Tool to convert an uploaded audio mixed with an image and generate a video format with image and uploaded audio in the video format.

Orchestrating ML Pipelines with Azure Data Factory

Leveraging Azure Data Factory for Scalable and Efficient Machine Learning Workflows

Developing a Astrology webapp version 1.

Initial version 1 of Astro app hosted at astroyuga.com

Improving the UI of this blogging app - V2

Changing the UI layout and improving the experience by modernizing the UI with custom styling

Leveraging CURSOR and Azure Services for Rapid Web Deployment

Accelerating Development and Deployment Cycles with AI Tools