Career in Fortune

About the Project

“Career in Fortune” is a career‐oriented platform designed to empower job seekers and professionals by combining job tracking, ATS (Applicant Tracking System) resume evaluation, and AI‐driven interview preparation under one roof. Inspired by gaps observed in traditional job portals—opaque ATS processes, fragmented resources, and minimal real‐time guidance—this project aims to streamline every step of a candidate’s journey, from crafting an optimized resume to practicing interviews tailored to data engineering, machine learning, and AI roles.

Inspiration

Personal Experience as a Job Seeker: While applying to data engineering and machine learning roles, I repeatedly encountered opaque ATS filters that would discard resumes without any transparent feedback. Candidates rarely understood why they got rejected, and it felt like shooting in the dark.
Fragmented Resources: There were many standalone interview-prep sites, resume scanners, and networking tools, but none that communicated with one another. I envisioned a unified platform where a user could track applications, receive instant ATS feedback, connect with recruiters, and practice domain-specific interview questions powered by AI.
Scaling Career Growth: Drawing on my background with Hadoop, Firestore, Vertex AI, and Google Gemini for NLP tasks, I wanted to build a system that aggregated Fortune 1000+ job listings, learned from user behavior, and proactively suggested next best actions—whether that meant refining keywords on a resume or drilling system-design mock interviews.

What It Does

Job Streak Tracking

Visualizes each user’s “application streak” (e.g., consecutive days of active searching or preparation).
Sends personalized reminders if a streak is about to break, helping maintain momentum.

ATS Resume Checker

Integrates a custom LLM (fine-tuned on Gemini AI) that simulates major ATS algorithms.
Provides users with an instant score breakdown: keyword alignment, formatting suggestions, and role-specific optimizations.

Application Pipeline Dashboard

Track Every Stage: From “Draft” and “Submitted” to “Interview Scheduled” and “Offer Received,” users can update and visualize their entire pipeline in real time.
Custom Tags & Notes: Annotate each job with role details, referral sources, application deadlines, and recruiter contacts.

Recruiter & Network Connection

In-Platform Messaging: Built on Firestore, users can connect with recruiters or peers directly via chat.
Automated Follow-Up Prompts: Schedule reminders to send thank-you notes, status check-ins, and networking follow-ups.

Interview & Exam Resources

AI-Driven Mock Interviews: Simulate both behavioral and technical interviews (data engineering, ML, AI) using Google Gemini’s conversational API.
Competitive Exam Prep Library: Curated question banks for GRE, SAT, and GMAT, organized by topic and difficulty—leveraging previous GRE/SAT/GMAT question sets I’ve developed for test preparation.

How We Built It

Frontend:
- Bolt.new + React: Started with Bolt.new for rapid prototyping of core pages, then migrated components into a React/TypeScript codebase.
- Tailwind CSS: Ensures a responsive, mobile-first design across desktop, tablet, and mobile viewports.
- Framer Motion & Lucide-React: Subtle animations (logo bounce, page transitions) and crisp icons make the UI feel polished and dynamic.
Backend & Database:
- Firebase Auth & Firestore: Manages user authentication, real-time messaging, user profiles, application logs, and interview transcripts.
- Firebase Functions: Hosts serverless endpoints that run ATS checks, send scheduled reminders via Cloud Tasks, and handle Stripe webhooks.
- Google Cloud (Vertex AI & Gemini AI):
- Resume Scoring Pipeline: Fine-tuned a Transformer model on Vertex AI using 1,200 anonymized resume–job pairs.
- Mock Interview Bot: Powered by Gemini’s conversational API, with domain-specific prompts for data engineering and AI roles.
- Netlify Hosting: Deployed static assets (React bundle) on Netlify for fast global CDN-backed delivery.
- Stripe Integration: Stripe Checkout manages B2C monthly subscriptions and B2B job-posting payments via Firebase Functions.
CI/CD & Infrastructure:
- GitHub Actions: Automates linting, unit tests, and deployment to Netlify on every commit to main.
- Terraform (GCP): Infrastructure as code for provisioning Firestore databases, Cloud Functions, IAM policies, and Vertex AI resources.
- Monitoring & Alerts: Google Cloud Monitoring tracks Cloud Functions errors, Firestore read/write quotas, and critical failures trigger Slack alerts.

Challenges We Ran Into

Scaling Real-Time Chat at Prototype Stage

Issue: As our beta usage grew from around 10 to 50 concurrent users, our single-region Firestore instance began experiencing increased latency in chat collections.
Solution: Migrated critical chat collections to multi-region replicas and implemented sharded counters to distribute write load. This ensured sub-100 ms latencies even under 50 concurrent connections, maintaining a smooth chat experience for early adopters.

Building a Reliable ATS Emulator

Issue: Every company’s ATS has its own quirks—different keyword density thresholds, resume parsing rules, and proprietary scoring criteria.
Solution: Collected 1,200 real resume–job pairs (anonymized), fine-tuned on Vertex AI, and added a rule-based layer to catch outliers (for example, PDF→Word conversion issues and image-based resumes).

Integrating Gemini AI Responsibly

Issue: Gemini occasionally hallucinated domain-specific technical answers during mock interviews, which risked misleading users.
Solution: Built a validation pipeline: every AI response is cross-referenced against an internal knowledge base (FAQs, official documentation). If confidence scores fall below a predefined threshold, the user receives, “I’m not 100 % certain—let me verify that for you,” instead of a potentially incorrect answer.

UI/UX for Static Scroll vs. Zoom-Out Deck

Issue: Our investor pitch deck’s zoom-out slider looked visually appealing but confused first-time users: scrolling often landed midway between sections.
Solution: Switched to a full-page snap-scroll implementation—each section snaps to full screen. Added scroll listeners in React that detect a user’s scroll intent and smoothly transition to the next or previous section, eliminating half-page views.

Balancing B2C and B2B Use Cases

Issue: Features like ATS scanning made sense for individuals but not necessarily for enterprise recruiting teams.
Solution: Modularized our codebase:
Client App (Job Seekers): Focused on job tracking, resume scoring, and interview prep.
Employer Portal (B2B): A separate Firestore collection group and React routes allow HR teams to post jobs, review ATS scores, and compare candidate pipelines. Permissions and billing logic differ for B2B, so we isolated those routes and functions accordingly.

Accomplishments That We’re Proud Of

Closed-Beta Sign-ups (May 2025): We invited 250 early adopters (friends, peers, and industry contacts) to our private beta. Within two weeks, 75 % of them completed their profiles, uploaded resumes, and tried the mock-interview flow.
Early User Feedback: From those 250 testers, we gathered 60 detailed feedback reports—most notably, suggestions on clarifying our ATS scoring explanations and requests for additional question categories in the mock-interview bot.
Prototype Stability: Despite running on a single-region Firestore instance, we maintained 99.8 % uptime for chat and dashboard features as concurrent users grew from 10 to 50. Sharded counters and optimized read/write patterns ensured sub-100 ms latencies even under load.
ATS Model Calibration: Using a dataset of 1,200 anonymized resume–job pairs, we fine-tuned our Gemini-based scoring model on Vertex AI. Early validation (5-fold cross-validation) showed 82 % alignment with real ATS outcomes—sufficient confidence to proceed with real-user testing.
Initial B2B Interest: Two mid-sized companies (not yet Fortune 1000) expressed interest in piloting our ATS analytics dashboard in exchange for feedback and anonymized hiring data once we move into the next closed-beta phase.

What We Learned

Data Collection & Anonymization Matter: Gathering real ATS acceptance/rejection pairs was more difficult than anticipated. We learned the importance of fast, automated PII removal and creating a continuous feedback loop so our model can retrain on new patterns as companies tweak their ATS criteria.
Simplicity vs. “Cool” UX: The snap-scroll interface felt impressive, but some users found it jarring—especially non-tech audiences. We realized a simpler static scroll with clearly labeled sections can be more intuitive. Always A/B test major UI patterns before committing.
AI Hallucinations Are Real: Gemini’s conversational abilities significantly speed up mock interviews, but hallucinations are unavoidable without guardrails. Building a fallback mechanism—“I’m not completely sure, let me look that up”—preserved user trust and prevented misinformation.
Monetization Nuances:

B2C: Users expect a low barrier to entry. Our \$1.99–\$2.99/month tier (with a 14-day free trial) hit the sweet spot.
B2B: Companies wanted volume discounts and analytics. Instead of charging only per job post, we introduced tiered subscriptions (\$19.99–\$29.99/month) for unlimited job postings and advanced ATS analytics.

Team & Budget Prioritization: With a \$2,000/person budget for existing employees and plans to hire three more, we learned how to prioritize features. We focused first on stabilizing chat and scaling Firestore before investing heavily in exam content and advanced analytics. Clear milestones helped us avoid scope creep.

What’s Next for Career in Fortune

Public Website Launch (June 2025):

Open the platform to the public with polished UI/UX.
Finalize mobile responsiveness and address all beta feedback.

Enhanced Interview Resources (Q4 2025):

Expand the question library to include behavioral case studies, system-design prompts, and domain-specific coding challenges.
Introduce live mock interviews, where users can schedule a 30-minute peer review session with feedback.

Mobile App & Push Notifications (Q1 2026):

Launch on iOS and Android, featuring real-time notifications for application deadlines, recruiter messages, and scheduled interviews.
Add offline support for resource articles, flashcards, and saved practice questions.

AI Interview Coach (Q2 2026):

Evolve from text-only mock interviews to voice-enabled practice using Gemini’s speech API. This will allow users to practice aloud and receive real-time pronunciation and tone feedback.
Build a “Performance Dashboard” tracking metrics such as “Answer Length,” “Keyword Coverage,” and “Soft-Skill Tone,” feeding back into a dynamic user score.

B2B Analytics & Resource Hub (Q3–Q4 2026):

Launch a dedicated employer analytics dashboard with:
- Pipeline Metrics: Time-to-hire, pipeline conversion rates, “ATS drop-off points.”
- Benchmark Reports: Compare a company’s ATS thresholds against industry averages.
Develop a “Candidate Matching Engine” that auto-ranks a shortlist based on skill-role fit, powered by BERT embeddings and custom clustering algorithms.

Partnerships & Continuous Learning:

Partner with online learning providers (Coursera, Udacity) to embed curated courses directly into our platform.
Automate certificate integration so users can display completed courses or bootcamps on their profiles, creating a more comprehensive candidate portfolio.