PathwayAI
Inspiration
It didn't start with a business plan or a whiteboard full of frameworks.
It started with a rejection email.
You know the one. The kind that arrives on a Tuesday afternoon when you've already refreshed your inbox seventeen times. "We regret to inform you..." — and somewhere in the middle of that polished corporate paragraph, you realize you spent four months preparing for a role you were never actually qualified for.
Not because you weren't smart enough. Not because you didn't work hard enough.
But because nobody ever told you exactly what the gap was.
Career advice for students is broken in a very specific, very quiet way. The internet tells you to "build your skills" and "tailor your resume." University career centers hand you PDF guides written in 2019. LinkedIn Premium shows you jobs you're already underqualified for. But nobody sits down with you and says: here is the exact distance between where you are and where you want to be — measured precisely, with a week-by-week plan to close it.
We kept asking: why does this tool not exist?
So we built it.
What It Does
Imagine having a brutally honest, incredibly smart career advisor available at midnight — one who doesn't give you generic advice, but actually reads the job postings you care about, studies your specific background, and tells you the truth.
That's PathwayAI.
You tell us your dream role and the companies you're targeting. While you're setting up your profile, PathwayAI is already working in the background — Tinyfish, our AI web agent, is crawling live job postings in real time. Not last month's listings. Not cached results. Actual openings, from actual companies, posted today. It reads them the way a hiring manager would, pulling out exactly what skills they're asking for, what experience they expect, and what responsibilities you'd own on day one.
Then you upload your resume. AWS Textract reads every line — your education, your work experience, your projects, your skills — and turns it into structured data we can actually analyze. You also take a 15-question assessment tailored specifically to your target role, so we understand not just what's on your resume, but how deeply you actually know it.
Then GPT-4o Mini does something no career advisor has ever done for you at scale: it compares your real background against real job requirements and gives you a number. A precise, honest match score:
$$\text{Match Score} = \frac{|\text{Your Skills} \cap \text{Required Skills}|}{|\text{Required Skills}|} \times 0.7 + \text{Assessment Score} \times 0.3$$
Not a vague "you're a good fit." Not a confidence boost. A real percentage that tells you exactly where you stand — and more importantly, exactly what's missing.
From that gap, PathwayAI builds your personalized week-by-week learning roadmap. Not a generic list of courses. A specific, sequenced plan built for you, for that role, based on those actual postings — so you know what to learn first, what can wait, and what will move the needle most.
Everything lives securely in InsForge — your profile, your gap report, your roadmap, your progress — so every time you come back, PathwayAI remembers exactly where you left off and how far you've come.
How We Built It
We started with a single HTML file and a dream. Seriously. The entire frontend began as pathwayai.html — one file, vanilla JavaScript, no frameworks. Just a vision and a text editor.
Then we layered in the real infrastructure piece by piece:
- InsForge — our entire backend lives here: authentication (Email, Google OAuth), PostgreSQL database, file storage for resumes, and Edge Functions that replaced our original Express server
- AWS Textract — real OCR-based resume parsing, not regex hacks
- Tinyfish — our web agent that reads live job postings the way a human would, extracting structured skill data from unstructured pages
- Redis Cloud — caching layer so job scans stay fast even under load
- GPT-4o Mini — the AI brain that compares resume vs. job requirements and generates the gap score and roadmap
- InsForge MCP + Cursor — we connected InsForge's MCP server directly into Cursor, which let us create database tables, deploy edge functions, and configure auth through natural language prompts — compressing days of work into hours The full flow:
User signs up via InsForge Auth
↓
Uploads resume → InsForge Storage
↓
Tinyfish scans live job postings
↓
AWS Textract parses resume
↓
GPT-4o Mini generates gap analysis
↓
Results saved to InsForge Database
↓
Personalized roadmap on dashboard
Challenges We Ran Into
The hardest challenge wasn't technical. It was emotional.
There's a very particular feeling that hits at 2am when the app you've been building for days still isn't working end-to-end. When the frontend looks beautiful but the backend is returning mock data. When you demo Screen 6 to a friend and secretly know that every single user is seeing the exact same hardcoded roadmap regardless of their resume. You smile and say "the real data integration is coming soon" — and then you go back and keep building.
Staying motivated through that gap between the vision and the working product was the real challenge.
On the technical side:
- The .env incident. Our AWS keys got committed to a public GitHub repo. Rotated within the hour — but those sixty minutes of watching the clock while revoking credentials and auditing access logs are not something we'll forget. We now check
.gitignorebefore every single push, without exception. - The TLS mystery. Redis Cloud requires TLS, but passing
tls: {}explicitly breaks the connection when the URL already starts withrediss://. The error was cryptic. The fix was removing one line. The debugging took three hours. That's software. - PDF has no structure. PDF.js gives you raw text, not structure. Turning a wall of extracted characters into cleanly separated Education, Experience, Projects, and Skills sections — using regex patterns and section header detection — taught us more about document parsing than any course ever could.
- Edge Functions aren't in the dashboard. InsForge Edge Functions are deployed via CLI, not the UI. Discovering this mid-build, while deep in integration, forced us to quickly learn a new workflow under real time pressure. We did. And it made us better.
Accomplishments That We're Proud Of
We are proud that it actually works.
A real student can go to pathwayai.insforge.site, sign up, upload their resume, and walk away with a personalized skill gap score and learning roadmap — powered by live job data, analyzed by AI — in under 60 seconds. That's not a prototype. That's not a mockup. That's a real product that real students can use tonight.
We're proud that we went from a single static HTML file to a fully integrated production stack — InsForge auth, database, storage, and edge functions; AWS resume parsing; Tinyfish live job scanning; GPT-4o Mini gap analysis; Redis caching — all connected, all working, all deployed.
We're proud of the match score formula. It's not a vibe. It's not a guess. It's a mathematically grounded comparison between a student's real skills and a role's real requirements, weighted by how they actually perform on the assessment.
But most of all — we're proud of what this could mean for someone who hasn't found us yet.
Somewhere out there, there's a student who's going to get a rejection email next Tuesday. They're going to feel exactly what inspired this project. And then maybe — maybe — they'll find PathwayAI. They'll upload their resume. They'll see their real score. They'll follow their roadmap. They'll close the gap.
And one day, instead of "We regret to inform you," they're going to get a different kind of email.
"We're thrilled to offer you the position."
That's what we're building toward. That's what we're proud of.
What We Learned
We learned that the hardest part of building is not the code — it's the decisions.
Every hour, there's a fork: this API or that one? Build this feature or cut it? Spend two hours polishing the UI or ship something that works? We learned to ship first and iterate fast.
We learned that every API has a secret the documentation doesn't tell you. You only find it by hitting the wall at 2am and reading the error message for the fifth time.
We learned that InsForge is genuinely powerful. Having auth, database, storage, and serverless functions in one platform — connected to your coding agent via MCP — is a developer experience that changes how fast you can build. What used to take days of config and deployment took us hours.
Most importantly, we learned that building for someone real changes everything. When we imagined a first-year student uploading their resume at midnight before a job application deadline — anxious, uncertain, hoping something would tell them they were ready — every design decision became clearer. Every bug fix felt urgent. Every feature had a human being behind it.
What's Next for Pathway AI
PathwayAI is live — but this is barely the beginning.
The most important conversation we had while building this wasn't with a developer or a mentor. It was with a career advisor. And what they told us stopped us in our tracks:
"The students who get jobs aren't always the most skilled. They're the ones who know how to network."
That conversation is shaping everything we build next.
🔗 Smart Networking — powered by your LinkedIn connections. Upload your LinkedIn data export and PathwayAI maps your entire network against your target role. We identify which of your existing connections are recruiters, hiring managers, or employees at your dream companies. Through Apollo.io integration, we find direct email addresses — so you can reach the right person with the right message at the right time, not just send applications into a void.
🎯 Recruiter Radar. We surface active recruiters who are currently hiring for your target role — people who are literally looking for someone like you right now — and give you a personalized outreach template based on your specific gap score and strengths.
📍 Events Near You. Networking isn't just digital. PathwayAI will surface relevant career fairs, industry meetups, and recruiting events happening in your city — so you can show up in person, prepared, with a clear story about who you are and what you're working toward.
📈 Live Progress Tracking. As you complete learning modules and build new skills, your match score updates in real time. You watch the gap close. Week by week. Point by point.
🏫 University Partnerships. Career centers can see aggregate skill gap data across their entire student body — which roles students are targeting, which skills are most commonly missing, which companies are most in demand — and adjust curriculum and resources accordingly.
The vision hasn't changed since day one:
No student should ever get a rejection email and not know why — and no student should ever feel like they have no path forward.
PathwayAI is that path.
Built with ❤️ using InsForge · AWS · Tinyfish · Redis Cloud · GPT-4o Mini Live at pathwayai.insforge.site
Built With
- amazon-web-services
- express.js
- github
- insforge
- javascript
- node.js
- redis
- tinyfish
Log in or sign up for Devpost to join the conversation.