
How I Built VidyaMitra: An AI Learning and Career Platform with Secure Resume Storage on Amazon S3
A student's story of building VidyaMitra, an AI career platform, and securing resume storage with a private S3 bucket, presigned URLs and AWS services. Built for the WeMakeDevs First Commit hackathon.
I Built a Career Platform for Students. Then I Made It Secure Enough to Trust With Their Resumes
How VidyaMitra uses 8 AWS services to turn a resume upload into a private, encrypted, AI-analyzed experience. Built for the WeMakeDevs x AWS First Commit hackathon.
The moment that changed my design
A resume is not a normal file. It has a person's full name, phone number, email, home city, and years of work history, all in one page.
When I first wired up resume uploads in VidyaMitra, the easy way was to drop files into a storage bucket, save the link in the database, and move on. It worked in five minutes.
Then I asked myself one question: "If someone guessed that link, could they read a stranger's resume?"
The honest answer was yes. So I threw away the easy version and rebuilt it properly. This post is the story of that rebuild, and how AWS made it much easier than I expected.
What is VidyaMitra?
I'm Gopal Mani, a Computer Science student at Mailam Engineering College in Chennai. Before college, I worked in tech support at Epson India and in Dubai. Every day I watched capable people get stuck because their tools were scattered and nothing connected.
Students preparing for jobs have the same problem. They use one site to learn, another to build a resume, another to check ATS filters, and another for career advice.
VidyaMitra puts all of that in one place:
Feature | What it does |
|---|---|
AI-assisted learning | Helps students understand topics and build skills |
Resume management | Upload and store resumes securely |
Smart Resume | Parses the resume and gives an ATS-style analysis |
Proctored assessments | Uses camera checks to keep practice tests honest |
Stack: React, Vite, TypeScript and Tailwind on the frontend, a backend API for logic and integrations, a database, the Gemini API for AI features, and AWS for the cloud layer.
The architecture at a glance
┌──────────────┐User ──► React ──► │ Backend API │ ──► Database └──────┬───────┘ │ ┌──────────┬───────┼────────┬───────────┐ ▼ ▼ ▼ ▼ ▼ Amazon S3 Lambda Rekognition SNS Secrets Manager (resumes) (Textract) (proctor) (alerts) + SSM Params │ CloudWatch + Cost Explorer
[ADD YOUR ARCHITECTURE DIAGRAM IMAGE HERE]
The key design rule: the frontend never touches AWS directly. Every request goes through my backend, which decides who is allowed to do what.
Building block 1: A private bucket with presigned URLs
This is the heart of the design, and the thing I'm proudest of.
The bucket:
- Fully private, with Block Public Access turned on
- Encrypted at rest with SSE-S3
- Files stored under user-specific paths, with the S3 details saved on the matching resume record
The upload flow:
- The student clicks upload.
- The frontend asks my backend for permission.
- The backend checks the student is logged in, then generates a presigned URL, a temporary link for one action on one file.
- The browser uploads straight to S3 using that link.
- The link expires.
Here's the core of the backend code, using the AWS SDK for JavaScript v3:
import { S3Client, PutObjectCommand } from "@aws-sdk/client-s3";import { getSignedUrl } from "@aws-sdk/s3-request-presigner";
const s3 = new S3Client({ region: process.env.AWS_REGION });
export async function createUploadUrl(userId: string, fileName: string) { // Each user only ever writes inside their own folder const key = `resumes/${userId}/${Date.now()}-${fileName}`;
const command = new PutObjectCommand({ Bucket: process.env.RESUME_BUCKET, Key: key, ContentType: "application/pdf", });
// Link works for 5 minutes, then it's dead const url = await getSignedUrl(s3, command, { expiresIn: 300 }); return { url, key };}
Downloads use the same pattern with
GetObjectCommand . The bucket stays private the whole time. Nobody gets in without a short-lived link that my backend approved.Building block 2: Turning a PDF into structured data
A private file is safe, but it's not useful until the app understands it.
When a resume lands in S3, an AWS Lambda function picks it up and uses Amazon Textract to extract the text. That text feeds the Smart Resume analysis, which checks things like keywords, structure and section order.
I liked this approach because Lambda only runs when a resume arrives. There's no server sitting idle, and I don't pay when nobody is uploading.
Building block 3: Keeping secrets out of my code
An early mistake I want to be honest about: at first I had keys and config values sitting in
.env files. That's fine for a laptop and risky for anything real.I moved them to:
- AWS Secrets Manager for sensitive values like API keys
- SSM Parameter Store for normal configuration
Now the code asks for secrets at runtime, and nothing sensitive lives in the repository.
Building block 4: Proctoring with Rekognition and alerting with SNS
For practice assessments, I wanted them to feel real. Amazon Rekognition analyzes camera frames for basic proctoring signals, such as no face in view or multiple faces. When something looks off, Amazon SNS sends a notification instead of the app silently ignoring it.
Building block 5: Watching the system and the bill
Two services keep me honest:
- CloudWatch for logs and metrics, so when an upload fails I can see why
- Cost Explorer so a student project doesn't turn into a surprise bill
As a student, watching cost isn't optional. It's part of the design.
The security stack, in one view
No single setting makes this secure. The layers work together:
Layer | What protects the resume |
|---|---|
Bucket | Private, Block Public Access on |
At rest | SSE-S3 encryption |
Access | Short-lived presigned URLs from the backend |
Identity | App-level authentication decides who can request a URL |
Paths | Each user's files live under their own prefix |
Secrets | Secrets Manager and SSM, never in code |
[ADD SCREENSHOT: S3 bucket permissions page showing Block Public Access ON] [ADD SCREENSHOT: S3 bucket encryption settings showing SSE-S3]
Try it: the resume flow
- Log in and open Resume.
- Upload a PDF. It goes straight to S3 through a presigned URL.
- Run Smart Resume.
- Get your analysis and improve the resume.
[ADD SCREENSHOT OR SHORT GIF OF THIS FLOW]
Results
[ADD REAL NUMBERS HERE. Examples of what to measure: upload time, Textract parse time, number of resumes tested, monthly AWS cost so far. Real numbers make the post far more convincing than adjectives.]
What I got wrong (and fixed)
- Public links first. My first version stored plain URLs. Presigned URLs fixed it.
- Secrets in
.env. Moving to Secrets Manager and SSM was worth the extra hour.
- No cost visibility. I added Cost Explorer checks before the bill surprised me.
- Trusting the frontend. The browser is never the authority. The backend is.
What I learned
- Security is layers, not a switch. Private bucket, encryption, expiring links and authentication each close a different gap.
- Serverless fits student budgets. Lambda runs only when needed, so cost stays low.
- Presigned URLs are underrated. They keep a bucket fully private while users still get smooth uploads.
- A feature isn't done when the button works. It's done when you know where the data goes and who can see it.
- Build real things. Reading about S3 is easy. Debugging a failed upload at midnight teaches you how it works.
What's next
[REPLACE WITH YOUR REAL PLANS. Only list things you actually intend to build.]
Final thoughts
VidyaMitra started as a simple frustration: students shouldn't need five tools to prepare for one career. It ended up teaching me how to handle other people's personal data responsibly in the cloud.
If you're a student thinking about adding AWS to a project, start with one real problem. Mine was "how do I store a resume safely?" and it led me through eight services and a lot of learning.
Thanks for reading. If you're building something similar, I'd love to hear how you approached it.
Gopal Mani is a Computer Science student and AI/LLM developer based in Chennai, India, a National Finalist at the OpenAI x NxtWave GenAI Buildathon, and an Oracle Cloud Infrastructure 2025 Certified Foundations Associate.
Tags:
aws amazon-s3 aws-lambda amazon-textract amazon-rekognition typescript security first-commit Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article