I ship AI systems the same way a good team ships software: scoped tickets, measured outcomes, no mystery work. Retrieval, agent orchestration, evaluation, and the product surfaces around them.
Details
Sprint stats
215
Repositories
606
Contributions
16
Repos contributed
9
Day streak
ScholarNotesPro — RAG research workspace
NotebookLM-style research app: PDF + YouTube ingestion, inline citations, and generated summaries, timelines, mind maps and study guides.
Daybreak — AI calendar scheduler
A terminal LangGraph scheduler rebuilt as a multi-user web app: per-user Google OAuth, SSE streaming, split-screen chat and agenda.
Persona chat — AI tech-educator simulator
Streaming chat that simulates the teaching style of two Hindi/Hinglish tech educators, with a persona toggle and prompt storage in Postgres.
aniketdev.blog — notebook book-blog
A book-style reading experience: markdown compiled to typed blocks, a page-flip reader, and a FastAPI + Neon backend with a markdown→blocks compiler.
Self-Consistency Answer Engine
Fans one prompt out to OpenAI, Claude and Gemini in parallel, then synthesizes a single answer with Claude — streamed live.
AI Expense Tracker (generative-ui)
Conversational expense tracker with generative UI, JWT auth and per-user data isolation on a persistent SQLite volume.
Live stock leaderboard (Redis)
Real-time NSE/BSE leaderboard on Redis sorted sets, with decoupled WebSocket write and button-click read layers.
“Ask my portfolio” chatbot
A small RAG agent that answers questions about my projects from this board’s own data and my write-ups.
Cost-aware model router
Routes each request to the cheapest model that clears a quality bar, with per-tenant budgets and live fallback.
Architecture decisions I've had to defend, written up after the fact. Each opens with the problem, the trade-off I took, and what it cost.
Unlike shared hosting, a VPS ships with nothing but the operating system. Everything else — firewall, users, web server, your app — is on…
Let's start simple and then move step by step toward more users. Relational and Non relational If many users access the system…
A load balancer distributes incoming network traffic across multiple servers to ensure that no single server is overloaded, and the traffic…
System design is the skill of choosing the right componts and putting them in right place to solve software engineering problem. Questions…
CRDT stands for Conflict free Replicated Data Type. It is a data structure designed for distributed systems where multiple replicas can…
Kafka is a distributed event streaming platform that allows multiple services to communicate through a durable, partitioned, and scalable…
Event driven architecture is a software architecture where services communicate by publishing and reacting to events instead of directly…
A cache stampede happens when many requests simultaneously discover that the same cached data is expired or missing and all try to rebuild…
Every cache miss is a potential load spike. When many requests miss the same cache entry at the same time, they can all hit the database…
A rate limiter controls how much traffic a client or service is allowed to send. In the HTTP world, that means capping the number of…