AI Agent Development
Build task-focused agents around model interaction, with clear workflow and execution boundaries.
For teams adding agent capabilities to a product.
Loading
Personal site / Engineering / Writing
Akshay Shinde
AI Engineer / Agentic AI Engineer
I build AI agents and production software, drawing on experience in software engineering, backend systems, and full-stack product development.
Building AI agents & production software
Skills
Next JS
React JS
Node JS
Express JS
Typescript
Prisma
Postgres
MongoDB
Mongoose
Mysql
Docker
AWS
CI/CD pipelines
Zod
Redux
Redux Toolkit
Recoil
JO
Jotai
Zustand
Redis
Agora
Socket.io
GraphQL
Hono JS
Three JS
Javascript
Tailwind CSS
Git
HTML
CSS
Jest
Vitest
OpenAI
Framer Motion
SEO
Drizzle
RPC
React Native
Cloudflare
Nginx
Python
Experience
D
Dautom
Senior Software Engineer
Jan 2025 - Present
F
Fanory
Software Engineer
May 2023 - Dec 2024
F
Freelance
Full Stack Engineer
Aug 2020 - Present
PI
Pegasus InfoCorp
Senior Software Engineer
Sept 2021 - May 2023
B
Bitmantics
Full Stack Developer
Sep 2019 - Sep 2021
A
Amdocs
Business and Technical Analyst
Jul 2018 - Aug 2019
About / Approach
Software engineering for agentic systems
I’m an AI Engineer and Agentic AI Engineer building AI agents and production software. I bring a software engineering foundation to AI work: the interfaces, APIs, data models, and system boundaries that turn a model capability into a product feature.
Engineering background
My background spans frontend and backend product development, API design, databases, architecture, and systems that run in production. I pay attention to how software behaves after it ships: reliability, performance, maintainability, and the operational context around it.
Current focus
My current focus is AI agents and agentic workflows: model interaction, streaming responses, retry behavior, and the boundaries needed to integrate AI into production software.
Builder mindset
I learn by building. I turn questions into working software, put ideas in product context, and refine the parts that fail or feel unclear.
What matters
Clear boundaries. Dependable behavior. Code that can be understood, changed, and operated by someone other than its author.
Engineering support
Engineering support
Practical engineering support for AI-enabled products and production software.
Agentic Workflow Engineering
Break a task into explicit steps and integrate model interactions with clear retry behavior and handoffs.
For processes that need more than a single model response.
AI Product Integration
Connect model interactions to product interfaces and backend systems, including streaming experiences.
For products integrating AI into an existing user journey.
Full-Stack Product Engineering
Build product features across frontend, APIs, and data layers with production behavior in mind.
For teams shipping or extending software products.
System Design and Architecture
Work through service boundaries, API and data models, performance needs, and implementation trade-offs.
For teams making consequential technical design decisions.
Technical Mentoring and Review
Review code and design choices, and reason through reliability and maintainability with your team.
For engineers who want a focused technical sounding board.
Talk through an engineering problem you’re working on.
Selected work / 01—05
Products, systems, and the details between them.
A working index of software across AI interaction, collaboration, publishing, and product infrastructure.
01 / 05
AI Software Engineer
AIOriginal
An agentic AI software engineering system designed to work through real software development tasks using specialized AI agents. The system orchestrates planning, research, implementation, testing, and review through a structured engineering workflow, with repository context, tools, verification, and feedback loops built into the process. The goal is to move beyond simple code generation toward an AI software engineer that can understand a codebase, plan changes, implement them, validate the result, and iteratively correct failures.
Built with
- TypeScript
- Node.js
- LangChain
- LangGraph
- Google Gemini
- AI Agents
- Agent Orchestration
- Tool Calling
02 / 05
Lifestack
WebAIOriginal
A personal life and productivity platform that brings everyday workflows into a single, structured workspace. Built as a modern Next.js application with server-rendered experiences across productivity, notes, calendar, nutrition, profile, and personal tools. The platform is designed as a foundation for multiple connected experiences, with a strong focus on useful workflows, thoughtful UX, and AI-powered capabilities rather than a collection of isolated features.
Built with
- Next.js
- TypeScript
- React
- Tailwind CSS
- shadcn/ui
- Server Components
- Framer Motion
- Server Actions
03 / 05
SENTINET-AI
Web
A sophisticated real-time AI model interaction platform with streaming capabilities, built with Next.js 15 and React 19.
- Real-time AI model interaction with streaming responses
- Live metrics tracking (tokens/sec, total tokens)
- Automatic retry logic with exponential backoff
- Configurable model parameters
- Progressive Web App (PWA) support
- Offline functionality
- Performance optimized with React memoization
- LaTeX equation support
- Dark/Light mode support
04 / 05
Taskenger (Trello)
WebAIOriginal
A full-stack web app to efficiently manage your tasks with priorities and rich, Trello-like features. Built with Next.js 14, Server Actions, Prisma, Stripe, Tailwind and MySQL.
05 / 05
CipherChat.ai
Web
You can send anonymous messages to your friends and family. And you can receive them as well. Additionally you can generate questions from AI for your suggestions. Built with Next js, Openai, mongo, tailwind, zod, resend, next auth and mongoose.
Notes / Technical writing
Thinking in systems.
- 01A Working Method for System DesignA repeatable sequence for designing a system: requirements, rough numbers, interfaces and data, the hard part, failure analysis, and the tradeoffs you are choosing. Worked through on a webhook delivery service.7 min read
- 02Distributed Systems Are About Partial FailureEvery remote call has three outcomes, not two. Idempotency, retry budgets, deadlines, quorums, and fencing tokens are what you build once you accept that.9 min read
- 03Graph Engineering: Modeling Agent Workflows as Explicit GraphsOnce an agent workflow has branches, parallel steps, approvals, and revision cycles, it is a graph whether or not you draw it. Typed state, nodes, conditional edges, reducers, checkpoints, and how to test and version the result.6 min read
Selected on X
Shorter thoughts.
- @ConsciousRideAI Engineering
As a backend engineer.
Please learn:
- One server-side language deeply (Node.js/TypeScript, Python, Java, Go - pick one and master it)
- API design & development (REST, GraphQL, gRPC, OpenAPI/Swagger, versioning, rate limiting)
Read the full post on X (post from Apr 15, 2026) - @ConsciousRideAI Engineering
90% of AI Engineer interviews in 2026 are just these 12 concepts repeated:
1. Prompt Engineering
• Zero-shot, few-shot, CoT, ToT, ReAct to steer LLMs
- Analogy: Giving the chef a super-detailed recipe with exact steps & plating instructions
Read the full post on X (post from Apr 15, 2026)
A little off the stack
Beyond code.
A few other places to find me, and a photo journal kept outside the work archive.
Last chapter / Contact
Get in touch.
Open to conversations about engineering opportunities, AI and agentic systems, product work, architecture, mentoring, and difficult technical problems.