Applytic

AI-powered job application tracker that learns from your rejections.

🚀 Start Using Applytic

Live in production. Free to use.

Applytic tracks every job application you submit, detects patterns across rejections (which resume version converts best, which source channel works, which company sizes respond), and uses Amazon Bedrock to turn that data into actionable coaching - delivered as a chat interface, tailored interview prep, rejection pattern alerts, and a weekly email digest.

Built end-to-end on AWS as a production-grade application. Every service is serverless, infrastructure is code, and every push auto-deploys via GitHub Actions.

Why I Built This

I was job hunting and had no data on why I was getting rejected. I had a spreadsheet with company names and "rejected" written next to most of them - but no signal on why. Was it my resume? The channel? The company size?

So I instrumented my own job search. Every application became a data point. After a few weeks I had enough data to see that my v1-generic resume had a 0% response rate from enterprise companies, while v3-ml-focused was getting interviews from startups via referrals. That's the kind of insight you can act on.

Features

Application tracking

Log applications with company, role, source channel, resume version, company size, job description URL, follow-up date
Kanban board with drag-and-drop status updates (Applied - Screened - Interview - Offer / Rejected)
Click any card to view full detail, edit all fields, change status, see timeline
Search by company/role, filter by source channel
Color-coded left border per status on kanban cards for instant visual scanning
Amber "Follow up" badge on overdue cards

Weekly goal tracking

Set a weekly application target on the Dashboard
Progress bar showing current week vs goal, turns green when met
Streak counter - consecutive weeks hitting your goal (fire icon)
Inline goal editing - click the pencil to update your target anytime

Follow-up reminders

Attach a follow-up date to any application
Daily email when applications are overdue for follow-up, grouped by user
Never miss a follow-up again

Notes timeline

Add timestamped notes to any application
Full per-application audit trail
Notes sorted oldest first for easy reading

CSV export and import

Export all applications to a CSV with one click - no Lambda, pure client-side
Import from CSV with full validation, preview, and error reporting
Download import template to get started quickly

AI insight engine

Pattern analysis across 6 dimensions: source channel, company size, resume version, role seniority, weekly velocity, status funnel
Response rate computed per bucket - shows exactly which resume version or source is working
AI coaching chat powered by Bedrock - answers questions like "why am I getting ghosted?" using your actual data as context
Markdown rendering in chat responses
Weekly email digest every Monday with stats + one AI-generated personalised tip

Interview prep

Generate 10 tailored interview questions per application using the job description URL and Amazon Nova Lite
Check off questions as you practice them and save your answer notes
Regenerate questions any time - only visible for applications in Interview or Offer status
Questions stored per application and persist across sessions

Rejection pattern alerts

Automatic pattern detection runs every Monday alongside the weekly digest
Detects: resume version with 0% response rate after 5+ applications, source channel with 0% response after 5+ applications, response rate dropping more than 20 points week-over-week
Amber alert banners on the Dashboard - dismiss with one click
Alerts also appear in the weekly digest email

Resume version tracker

Upload multiple PDF versions to S3 via presigned URLs
Tag each application with which version was used
Analytics shows conversion rate per version side-by-side

UI / UX

Full dark mode with system preference detection, persisted to localStorage
Mobile responsive - hamburger sidebar on small screens
Loading skeletons on every page instead of blank states
Meaningful empty states with calls to action
Toast notifications (top-center)

Architecture

Key Engineering Decisions

Single-table DynamoDB - two entity types (APPLICATION, STATUS_EVENT) in one table with composite keys. Every query is O(1). No joins. Scales to millions of requests/sec with no configuration changes.

Pattern analysis before LLM - the insights Lambda computes structured metrics (response rates per bucket) before calling Bedrock. The LLM receives hard numbers as context, not raw records. This makes advice data-driven and specific rather than generic.

Serverless-first - no EC2, no containers, no idle servers. Lambda + API Gateway + DynamoDB means ~$0 at low volume and automatic scaling. EventBridge replaces cron servers entirely.

ARM64 Lambda - all functions run on Graviton2 (ARM64) for ~20% cost reduction and faster cold starts vs x86.

Presigned S3 URLs for resume upload - the Lambda generates a presigned URL and returns it to the client. The file then uploads directly from the browser to S3 - never passes through Lambda.

Lambda Layer for shared code - single source of truth for CORS, auth extraction, error formatting, Pydantic validation, and X-Ray tracing. All seven Lambda handlers share it.

Client-side CSV - export and import are handled entirely in the browser. No Lambda invocation needed for either operation, keeping them fast and free.

Streak computation - looks back 8 weeks maximum, current week excluded (may be incomplete). Walks back week by week, breaks on first miss.

Tech Stack

Layer	Service
Frontend	React 18, TypeScript, Vite, Tailwind CSS, react-markdown
Auth	Amazon Cognito (email + JWT)
API	API Gateway REST + Lambda (Python 3.12, ARM64)
AI / ML	Amazon Bedrock - Amazon Nova Lite
Database	DynamoDB - single-table design, PAY_PER_REQUEST
Storage	S3 - resume versioning + frontend hosting
CDN	CloudFront (450+ edge locations)
Scheduling	EventBridge cron (Monday 8am UTC digest, daily 9am UTC follow-ups)
Email	Amazon SES
IaC	AWS CDK v2 TypeScript
CI/CD	GitHub Actions

Project Structure

applytic/
├── cdk/                    # AWS CDK v2 stack - all infrastructure as code
│   ├── bin/app.ts
│   └── lib/applytic-stack.ts
├── lambdas/                # Python 3.12 Lambda handlers
│   ├── applications/       # CRUD + presigned S3 URL generation
│   ├── insights/           # Pattern analysis + Bedrock AI coaching
│   ├── digest/             # Weekly SES email digest
│   ├── followup/           # Daily follow-up reminder emails
│   ├── settings/           # Weekly goal + streak tracking
│   ├── notes/              # Per-application notes timeline
│   ├── cognito_verify/     # Post Confirmation trigger - SES email verification
│   └── shared_layer/       # Lambda Layer - shared middleware, Pydantic, X-Ray
├── frontend/               # React + Vite + Tailwind CSS
│   └── src/
│       ├── components/     # Kanban, Analytics, Chat, Sidebar, Resume, CSV
│       ├── hooks/          # useApplications, useSettings, useNotes
│       ├── lib/            # API client, Amplify config, theme, csv utils
│       ├── pages/          # Dashboard
│       └── types/          # Shared TypeScript types
├── tests/                  # 207 pytest tests (90.75% coverage)
│   ├── test_applications.py
│   ├── test_applications_integration.py
│   ├── test_insights.py
│   ├── test_insights_integration.py
│   ├── test_digest.py
│   ├── test_cognito_verify.py
│   ├── test_followup.py
│   ├── test_settings.py
│   └── test_notes.py
├── scripts/
│   ├── seed_data.py
│   ├── build_layer.sh
│   └── setup_oidc_v11.sh
└── .github/workflows/
    ├── deploy.yml
    └── codeql.yml

DynamoDB Single-Table Design

Table name: applytic

Entity	PK	SK	GSI1PK	GSI1SK
Application	`USER#userId`	`APP#appId`	`USER#userId`	`DATE#timestamp`
Status event	`APP#appId`	`EVENT#timestamp`	-	-
Rate limit	`RATELIMIT#userId`	`DATE#YYYY-MM-DD`	-	-
User settings	`USER#userId`	`SETTINGS`	-	-
Note	`APP#appId`	`NOTE#timestamp#noteId`	-	-

Access patterns:

List all applications for user - GSI1 query on USER#userId sorted by date
Get single application - Main table get on USER#userId + APP#appId
Get status history for an app - Query APP#appId with EVENT# prefix
Get notes for an app - Query APP#appId with NOTE# prefix
Get user settings - Get on USER#userId + SETTINGS

API Routes

GET    /v1/applications
POST   /v1/applications
GET    /v1/applications/{appId}
PUT    /v1/applications/{appId}
DELETE /v1/applications/{appId}
POST   /v1/applications/{appId}/status
GET    /v1/applications/{appId}/notes
POST   /v1/applications/{appId}/notes
DELETE /v1/applications/{appId}/notes/{noteId}
POST   /v1/resumes/upload-url
GET    /v1/resumes/list
GET    /v1/insights
POST   /v1/insights/chat
GET    /v1/users/settings
PUT    /v1/users/settings
POST   /v1/applications/{appId}/interview-prep/generate
GET    /v1/applications/{appId}/interview-prep
PUT    /v1/applications/{appId}/interview-prep/{questionId}
GET    /v1/users/alerts
PUT    /v1/users/alerts/{alertId}/dismiss

All routes protected by Cognito JWT authorizer.

Scale and Cost

Metric	Value
API throughput	10,000 req/sec (API Gateway default)
DynamoDB SLA	99.999% availability
Lambda cold start	~300-400ms (Python 3.12 ARM64)
CloudFront edge locations	450+ worldwide
Cost at 0 users	~$0/month
Cost at 100 users	~$2-5/month
Cost at 1,000 users	~$15-30/month
Backend test suite	410 tests, 70% coverage threshold
Frontend test suite	31 Vitest tests
Full CDK deploy from scratch	under 3 minutes

CI/CD Pipeline

push to main
    ├── test (pytest 410 tests, 70% coverage threshold)
    ├── test-frontend (vitest 31 tests)
    └── (both must pass before)
            ├── deploy-backend (cdk deploy)
            ├── deploy-frontend-aws (s3 sync + cloudfront invalidation)
            └── deploy-frontend-pages (GitHub Pages)

Local Setup

Prerequisites

Node.js 18+
Python 3.12
AWS CLI configured (aws configure)
AWS CDK CLI: npm install -g aws-cdk

Deploy backend

bash scripts/build_layer.sh
cd cdk && npm install && cdk deploy

Run frontend locally

cd frontend && npm install && npm run dev

Run tests

pip install pytest boto3 pydantic aws-lambda-powertools pytest-cov
pip install "moto[dynamodb,s3,ses,cognitoidp]"
python -m pytest tests/ -v --tb=short --cov=lambdas --cov-fail-under=70

Run frontend tests

cd frontend && npm run test

Seed demo data

cd scripts && python seed_data.py --user-id YOUR_COGNITO_SUB

Issues Encountered and Fixed

1. TypeScript error on CDK deploy (`Partial<FunctionProps>` type)

Issue: runtime became optional when using Partial<FunctionProps>, causing TypeScript to reject undefined values.
Fix: Added runtime explicitly to each Lambda and removed the shared defaults spread.

2. esbuild crash on Windows

Issue: esbuild failed due to corrupted dependencies or a Node.js version mismatch.
Fix: Deleted node_modules and package-lock.json, reinstalled dependencies, and used Node 18/20 LTS.

3. pytest module name collision

Issue: Multiple Lambda handlers were named handler.py, causing import collisions during tests.
Fix: Used importlib.util.spec_from_file_location() with unique module names.

4. Timezone-aware datetime bug in velocity calculation

Issue: dateApplied was stored as a YYYY-MM-DD string and could not be subtracted from a timezone-aware datetime.
Fix: Detected 10-character date strings and added UTC timezone information before subtraction.

5. Chat history lost on navigation

Issue: Component-level useState was reset whenever the component unmounted.
Fix: Lifted message state to App.tsx and passed it down as props.

📋 Check out all the issues faced and solved: Issues_Encountered_and_Fixed.md

Changelog

See CHANGELOG.md for full version history.

Contributing

See CONTRIBUTING.md for setup instructions, branching conventions, and the PR process.

Author

Hardik - Github

Name		Name	Last commit message	Last commit date
Latest commit History 104 Commits
.github		.github
api-docs		api-docs
api		api
cdk		cdk
frontend		frontend
lambdas		lambdas
scripts		scripts
tests		tests
.coverage		.coverage
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
Issues_Encountered_and_Fixed.md		Issues_Encountered_and_Fixed.md
LICENSE		LICENSE
README.md		README.md
applytic.gif		applytic.gif
architecture.png		architecture.png
package-lock.json		package-lock.json
pytest.ini		pytest.ini

Navigation Menu

Repository files navigation

Applytic

🚀 Start Using Applytic

Why I Built This

Features

Architecture

Key Engineering Decisions

Tech Stack

Project Structure

DynamoDB Single-Table Design

API Routes

Scale and Cost

CI/CD Pipeline

Local Setup

Prerequisites

Deploy backend

Run frontend locally

Run tests

Run frontend tests

Seed demo data

Issues Encountered and Fixed

1. TypeScript error on CDK deploy (`Partial<FunctionProps>` type)

2. esbuild crash on Windows

3. pytest module name collision

4. Timezone-aware datetime bug in velocity calculation

5. Chat history lost on navigation

Changelog

Contributing

Author

About

Releases

Packages

Contributors

Languages

Folders and files

Latest commit

History

Repository files navigation

Applytic

🚀 Start Using Applytic

Why I Built This

Features

Architecture

Key Engineering Decisions

Tech Stack

Project Structure

DynamoDB Single-Table Design

API Routes

Scale and Cost

CI/CD Pipeline

Local Setup

Prerequisites

Deploy backend

Run frontend locally

Run tests

Run frontend tests

Seed demo data

Issues Encountered and Fixed

1. TypeScript error on CDK deploy (Partial<FunctionProps> type)

2. esbuild crash on Windows

3. pytest module name collision

4. Timezone-aware datetime bug in velocity calculation

5. Chat history lost on navigation

Changelog

Contributing

Author

About

Topics

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

1. TypeScript error on CDK deploy (`Partial<FunctionProps>` type)