FULL-STACK DEVELOPER · B.TECH CSE · BATCH OF 2027
Full-Stack Developer
Building Web Apps
With AI Features
I got into AI the unglamorous way: debugging a RAG pipeline that kept confidently citing sources that didn’t exist, and realizing I cared more about fixing why it broke than about the demo it was supposed to power. That’s pretty much how I approach full-stack work too: I’d rather ship something boring that holds up under real traffic than something flashy that only survives the first click.
A Bit About Me
I started out doing what most CS students do: competitive programming, MERN projects, the usual DSA grind. The pull toward AI happened almost by accident: I plugged an LLM into a side project expecting it to just work, and instead spent a weekend learning why it didn’t. Bad retrieval, no evaluation, no idea when the model was confidently wrong. That was more interesting to me than the model itself.
So now I sit at the intersection on purpose: full-stack engineering for the foundation, AI systems for what I want to build on top of it. I’m opinionated about this: an AI feature bolted onto a shaky backend is still a shaky product. I’d rather spend an extra day on error handling and session logic than ship an AI demo that looks great until someone actually uses it wrong.
Understanding AI systems, to me, isn’t about collecting certificates. It’s knowing that a RAG pipeline can be technically correct at every step and still return garbage, and caring enough to go find out why.
A tradeoff I’d make again
On ReputeChain I could’ve prototyped against a mocked chain to move faster. I used the real testnet from day one instead, which meant gas costs and confirmation delays showed up in week one instead of surprising me in week six. Slower start, fewer surprises later. I’ll take that trade most days.
AkshatProjects
Four projects. Open “Behind the build” on a card to see what was hard and the trade-offs I made.
01 // Blockchain / Decentralized Identity
ReputeChain
A reputation platform on Ethereum (testnet). Smart contracts record credibility scores on-chain so anyone can verify them, instead of trusting a central database.
02 // Competitive Programming / Real-time Systems
SyncRank
A leaderboard for a group of competitive programmers. It pulls submissions from LeetCode and Codeforces and turns them into one shared ranking.
What I learned
Codeforces and LeetCode don't rate-limit the same way, and neither tells you nicely when it does. I learned to build the sync job assuming both feeds will occasionally lie or go quiet, and to make the leaderboard degrade gracefully instead of freezing on a bad poll.
— Went with polling instead of webhooks. Neither platform offers reliable push events, so I built a jittered polling schedule instead.
— Stored raw submission snapshots even when unchanged, which cost more storage but made debugging sync drift far less painful.
Sync interval
45–90s
Sources
LeetCode, Codeforces
03 // Developer Tooling / Authentication
Open_authtool
An authentication layer for Node/Express apps: OAuth 2.0 sign-in, JWT access and refresh tokens, sessions, and permission checks, with a small set of defaults.
04 // Full-Stack Project
Mainstage
A movie platform. The write-up is still in progress; the code is on GitHub.
Skills
- Python
- C/C++
- JavaScript
- TypeScript
- SQL
- React.js
- Next.js
- TailwindCSS
- Node.js
- Express.js
- REST APIs
- MongoDB
- MySQL
- Google Gemini AI
- Vercel AI SDK
- LlamaIndex
- RAG
- LangChain
- DSA
- System Design
- Authentication
A Short, Real Timeline.
Work, projects, certifications and education in one timeline.
GOVT OF MP INITIATIVE
AI Trainer, MP Government EQIP Programme
Madhya Pradesh Government
Selected as an AI Trainer. Trained about 30 government school teachers a day over a 6-day programme on AI tools and smart learning methods.
PROJECT
Blockchain / Decentralized Identity
A reputation platform on Ethereum (testnet). Smart contracts record credibility scores on-chain so anyone can verify them, instead of trusting a central database.
PROJECT
Competitive Programming / Real-time Systems
A leaderboard for a group of competitive programmers. It pulls submissions from LeetCode and Codeforces and turns them into one shared ranking.
PROJECT
Developer Tooling / Authentication
An authentication layer for Node/Express apps: OAuth 2.0 sign-in, JWT access and refresh tokens, sessions, and permission checks, with a small set of defaults.
PROJECT
Full-Stack Project
A movie platform. The write-up is still in progress; the code is on GitHub.
CERTIFICATION
Infosys Springboard
Introduction to Artificial Intelligence, Introduction to Deep Learning and Computer Vision 101.
CERTIFICATION
Kaggle Learn
Micro-courses on data science and machine learning workflows.
EDUCATION
B.Tech, Computer Science & Engineering
CurrentKCC Institute of Technology and Management, Greater Noida (Affiliated to AKTU)
B.Tech in Computer Science & Engineering, alongside the projects and courses above.