The tools we trust. And why.

React — The baseline for production web apps. Server components, deep ecosystem, team-wide fluency.

Next.js — Routing, caching, and server components wired right out of the box. Right default for most apps.

TypeScript — Every build is typed. Catches whole classes of bugs before runtime and makes refactors safer.

Astro — Content-first pages that ship minimal JS. Islands architecture where interactivity is needed.

Vite — Fast dev server and build pipeline. Where speed of iteration matters more than a full framework.

TanStack — Query, Router, Table — the typed primitives we reach for when the app gets state-heavy.

Node.js — Mature ecosystem, great primitives for I/O-bound services, and the language we share with the frontend.

Bun — Faster JS runtime with a batteries-included toolkit — install, bundle, test — in one binary.

Hono — Tiny, fast web framework for edge runtimes and Bun. Great for APIs that need to be close to the user.

NestJS — Opinionated, modular Node framework for larger services where structure and DI pay off.

Express — The minimal, no-ceremony choice when all we need is a small HTTP layer.

GraphQL — Strongly typed API contract when the client needs fine-grained control over the response shape.

Python — Default for AI, data, and ML workloads — rich scientific stack and mature tooling.

Go — When latency and throughput matter. Strong concurrency primitives, minimal runtime.

OpenAI — GPT-4 class models for production features — robust function calling, structured output, solid latency.

Anthropic — Claude for reasoning-heavy, agentic, and long-context workloads.

Gemini — Google Gemini for multi-modal tasks and tight integration with Google Cloud tooling.

Groq — Ultra-low-latency LLM inference when response time is the UX.

Mistral — Open-weight and hosted models when we need cost-performance control.

Hugging Face — Open-source model hub, datasets, and inference endpoints for research-grade AI.

Cohere — High-quality embeddings and rerankers for retrieval — a strong default for RAG quality.

Replicate — Run open-source models on demand without operating GPUs ourselves.

LangChain — For chaining LLM calls, tool use, and memory when we need a well-trodden abstraction.

LangGraph — Graph-based agent flows with proper state, checkpoints, and human-in-the-loop.

LlamaIndex — Retrieval pipelines, indexing, and data-aware LLM apps.

CrewAI — Coordinating multiple specialised agents on a shared goal.

Vercel AI SDK — Typed React / Node helpers for streaming, tool calls, and chat UIs in Next.js apps.

Pinecone — Managed vector store for retrieval-augmented generation and semantic search at scale.

Qdrant — Self-hostable vector search when data has to stay inside your infrastructure.

Chroma — Embedded-first vector store for lightweight RAG prototypes and small corpora.

Ollama — Running models locally for privacy-sensitive workloads and dev-time iteration.

LangSmith — Tracing, evaluation, and debugging for LLM apps in production.

Langfuse — Self-hostable alternative for LLM tracing and evals when data residency matters.

PyTorch — When a task truly needs a custom-trained or fine-tuned model, not a prompt.

PostgreSQL — Battle-tested, feature-rich, right for most business applications.

Drizzle — Typed SQL in TypeScript — edge-runtime friendly, no hidden query magic.

MongoDB — For unstructured or rapidly evolving data schemas.

Redis — Caching, session management, and real-time pub/sub. Keeps apps fast under load.

Supabase — Postgres with auth, storage, and real-time built in. Accelerates smaller teams.

AWS — Default cloud for larger deployments. Deep service catalogue, proven reliability.

Google Cloud — Where AI and data workloads benefit from native tooling (Vertex, BigQuery).

Docker — Consistent environments from dev to prod. The packaging baseline.

Kubernetes — Orchestration at scale — only when the complexity actually earns its keep.

Vercel — The fastest path to production for Next.js and static apps. Great DX.

GitHub Actions — Automated testing and deployment pipelines on every project, no exceptions.

Make — Visual automation workflows where custom code is overkill.

Zapier — Wide app coverage for quick, reliable glue between SaaS tools.

n8n — When automations need to stay inside your infrastructure.

HubSpot — For full-funnel CRM integrations and lifecycle marketing.

Salesforce — Enterprise sales stacks and custom Salesforce API work.

Stripe — Recurring billing, invoicing, and global payment flows.

Razorpay — UPI, cards, and India-first payment flows with clean APIs.

Twilio — SMS, voice, and programmable messaging at scale.

WhatsApp Business — Customer conversations where WhatsApp is the primary channel.