I started my career as an intern at an early-stage startup while I was still in college. In a small team, I had the opportunity to work across different parts of the product and take on whatever needed to be built. That experience helped me discover that software engineering was something I genuinely enjoyed.
I continued learning full-stack development independently and gradually took ownership of the product end to end. After receiving a pre-placement offer, I became the startup's sole full-time engineer, responsible for turning ideas into reliable, production-ready systems.
Today, I'm a Full Stack Developer with 7+ years of experience building real-time, data-intensive products across Web2 and Web3. My work spans TypeScript, Node.js, React, NestJS, Next.js, Python, FastAPI and smart-contract integrations. I enjoy solving complex problems, improving performance, and owning features from architecture through production.
Outside of work, I enjoy traveling, riding my motorbike, driving, and meeting people from different cultures.
// experience
Where I've worked
// projects
Projects
A closer look at the systems, product decisions, and engineering trade-offs behind the work.
AI DEVELOPER TOOLS
01
AI Code Reviewer
A multi-provider AI code-review workspace that streams structured findings, security issues, suggestions, and quality scores while keeping provider keys under the user’s control.
CONTEXT
A full-stack Next.js application where developers paste code, select a language and model, and receive a progressively rendered review from Claude, GPT-4o, or a local Ollama model.
THE PROBLEM
AI-generated reviews are difficult to render and validate when providers return inconsistent plain text. The application needed a predictable review schema, provider isolation, secure BYOK handling, and protection against abuse in public deployments.
WHAT I BUILT
→Used the Vercel AI SDK to stream review results progressively instead of waiting for the complete response.
→Defined structured validation for summaries, quality scores, bugs, security findings, suggestions, and positive observations.
→Implemented Claude, GPT-4o, and Ollama behind a common provider contract.
→Added provider-specific key handling so keys remain isolated and are not persisted in local storage or cookies.
→Added hosted-environment controls for Ollama and Upstash Redis rate limiting for public deployments.
The application provides progressive feedback during review generation while keeping output predictable across providers. The deployed version supports BYOK without requiring Anthropic or OpenAI server keys, and the project is covered by 13 test suites with 114 passing tests.
Next.jsTypeScriptVercel AI SDKClaudeGPT-4oOllamaZodUpstash RedisJest
A financially correct portfolio-order allocation API that preserves totals despite rounding and fractional-share constraints.
CONTEXT
A NestJS and TypeScript proof of concept for a wealth-management platform. The service converts a model portfolio and investment amount into executable stock orders.
THE PROBLEM
Naive floating-point calculations can overspend, lose cents, or make fractional-share discrepancies impossible to explain. The system needed to preserve allocations, avoid overspending, and schedule execution around US market hours.
WHAT I BUILT
→Separated targetAmount, executableAmount, and unallocatedCash so every rounding decision is explicit.
→Used decimal arithmetic and the largest-remainder method to reconcile allocations down to the cent.
→Added strict DTO validation, duplicate-symbol detection, fractional-share truncation, and BUY/SELL order support.
→Implemented market-hour scheduling, domain events, Swagger documentation, unit tests, E2E tests, and Docker support.
The API preserves two financial invariants: sum(targetAmount) equals totalAmount, while sum(executableAmount) plus unallocatedCash also equals totalAmount.
Built a real-time perpetual trading experience integrating dYdX and GMX v2, with live positions, order management, PnL tracking, and on-chain transaction flows.
CONTEXT
Worked as a Full Stack Engineer on Rage Trade’s perpetual trading platform, contributing across the React/Next.js frontend, backend APIs, and Web3 integration layer.
THE PROBLEM
The application had to present a consistent trading experience while combining live market data, wallet state, protocol data, and asynchronous blockchain transactions. Users needed clear visibility into open positions, orders, PnL, pending transactions, and confirmed state.
WHAT I BUILT
→Built real-time interfaces for open positions, orders, PnL, and market data.
→Integrated dYdX and GMX v2 protocol interactions into the trading flow.
→Designed transaction states for wallet confirmation, submission, pending execution, confirmation, and failure.
→Handled synchronization between on-chain protocol state, market data, and client-side UI state.
→Owned features across frontend, backend API integration, and production delivery.
ARCHITECTURE
Wallet and protocol integrations → market-data and backend APIs → normalized client state → positions, orders, and PnL interface
OUTCOME
Delivered a responsive trading interface that made protocol interactions and transaction state understandable to users while keeping positions, orders, and PnL synchronized with live data.
Real-time data systemsKafkaPerformance OptimizationCaching
Database
PostgreSQLMongoDB
Web3
Smart Contracts IntegrationwagmiEthers.js
Cloud & Infrastructure
AWSGCPDocker
Testing
JestCypress
APIs & Docs
OpenAPI (Swagger)Mintlify
// testimonials
What people say
"
Working with Shubham was one of the best experiences. His work ethic and dedication to figuring out new challenges and taking it to the finish line were super impressive. He is a team player and I'm sure he'll add lots of value to any team he decides to work with in the future.
The "Fullstack" Guy, that's what I call him. Worked on a very complex frontend application which is tightly coupled with a Salesforce platform via Django which involves high volume of sensitive data flow to and fro.
He was involved in designing software platforms, responsive UI design and web page design and development. He played a significant role, often, in developing and implementing overall web strategies, and dealt with the details along with the bigger picture. His coding was excellent and his designs were unique and innovative.
A huge part of his job with us was coordinating interdepartmentally to facilitate the completion of company-wide goals through web development.
He was always thoughtful and considerate, an excellent team player who can also work independently and do a fantastic job.
Shubham can seemingly do anything, so he’s very highly recommended for a full stack engineer role at any firm.
— Ashok Kumar Nayak
Director of Technical Architecture (Salesforce), Cint
Shubham is very sincere and hard working. once given direction on what needs to be done, he was able to figure out how to do it - he almost taught himself the entire full stack development despite very limited technical mentoring bandwidth in the company. also, took great ownership and interest in the company's progress..