Practice

AI agents & workflows

Agents that carry real tools and are held to real checks — reading a codebase, writing and reviewing changes, and running inside a workflow rather than beside it. The interesting part is never the prompt; it is the control flow, the grounding, and the evidence that it worked.

Covers

• Claude and GPT, plus open-source models
• Tool use, function calling and MCP servers
• LangGraph for state, branching and recovery
• Knowledge graphs and retrieval for grounding
• Evals, tracing and guardrails before shipping
• Wired into CI/CD, Git and the editor

Backend

Services designed for the day they are under load: clear boundaries, migrations that run clean, and observability wired in before it is needed. The substrate an agent has to run on if it is going to run at all.

Covers

• NestJS, Express, Actix Web, Gin
• Microservice architecture and EC2 deployment
• Payment integration (Razorpay, PayU)
• REST and GraphQL
• Distributed systems

Full stack

One person holding the whole line — interface, service, model and chain — so the seams between them stop being where the bugs live.

Covers

• Frontend development
• Backend development
• Agent and model integration
• Cross-platform with React Native
• Ongoing developer support

Frontend

Interfaces that stay legible as the product grows: typed components, honest loading and error states, and animation used to explain rather than to decorate. Streaming and tool-call surfaces included — an agent that cannot be watched cannot be trusted.

Covers

• React and Next.js, app-router first
• Streaming responses and tool-call UI
• Tailwind CSS and styled-components
• REST, GraphQL and TanStack Query
• React Spring and Three.js

Blockchain

Contracts written to be read by an auditor, and the indexing and decoding around them that turns on-chain noise into something an application — or an agent — can trust.

Covers

• Solana, Ethereum, BSC, Polygon, Bitcoin
• Contract development and deployment
• Contract testing
• Indexing and on-chain research
• Agent identity and payments on-chain

Consulting

Reading a system and saying plainly what it needs: the architecture, the sequencing, and — increasingly — which parts genuinely want a model and which parts are better off without one.

Covers

• Where agents help, and where they do not
• Architecture and sequencing
• Research and development
• Resourcing
• Developer support