NexSetu

Landing page
AI parser
Recruiting page
Candidate Profiles
Profile setup
Job Postings
Upskilling Flow

Inspiration

The inspiration for NexSetu came from a striking contradiction we noticed in today's job market. When BlackRock's CEO Larry Fink announced that the U.S. needs 500,000 more electricians to support AI data centers and infrastructure projects, something clicked for us.

While the tech world obsesses over AI replacing jobs, there's this massive irony: AI can't touch blue-collar work, yet companies are desperately struggling to find these workers fast enough. We started researching and discovered that:

•⁠ ⁠Hiring Lag: Companies take 60-90 days to hire 100 skilled workers through traditional methods.

•⁠ ⁠Cost of Delay: Every week of delay costs them $50,000 to $200,000 in project delays.

•⁠ ⁠Worker Loss: Electricians sit unemployed for 6-10 weeks between gigs, losing $8,000–$15,000 annually.

•⁠ ⁠Agency Fees: Staffing agencies extract 30–50% of what companies pay, leaving workers with only 70% of their value.

The real kicker? Jensen Huang said blue-collar jobs will remain untouched by AI for decades. So we thought: what if we use AI not to replace these workers, but to empower them? That's how NexSetu was born—a nexus (connection point) that bridges skilled workers to infrastructure opportunities.

What it does

NexSetu streamlines the hiring and career growth process through four key pillars:

•⁠⁠Instant Requirement Extraction: Companies paste a job description and our AI extracts requirements in 5 seconds—no forms to fill out.

•⁠ ⁠Smart Match Scores: We show ranked candidates with match scores. A 95% score means they are ready to go; a 78% score means they are missing one certification, and we show exactly how to get it in 2 weeks.

•⁠⁠AI Upskilling Roadmaps: We generate roadmaps like: "Get Arc Flash certification for $750, qualify for 45 jobs paying $15K more per year."

•⁠Bulk Hiring Tools: Employers can select 50 candidates and send interview invites with one click. What took 3 months now takes 3 weeks.

How we built it

Built collaboratively during SwampHacks 2026:

•⁠ ⁠Backend: FastAPI with SQLAlchemy ORM. AI logic is centralized in a dedicated ⁠ ai_service.py ⁠.

•⁠ ⁠Frontend: React + TypeScript SPA with Vite for high-performance UI/UX.

•⁠ ⁠AI Integration: Custom LLM prompt engineering to transform unstructured job data into validated Pydantic schemas.

Challenges we ran into

1. Pricing Model Uncertainty

We settled on a hybrid model to ensure sustainability and worker accessibility:

•⁠ ⁠Workers: Free forever.

•⁠ ⁠Employers: $2,000/month + 5% placement fee.

The Revenue Formula: $$Annual Revenue per Client = (12 \times \$2,000) + (0.05 \times Avg Placements \times Avg Salary)$$

2. Competing With "Good Enough" Solutions

We realized our true competitors aren't job boards like LinkedIn, but staffing agencies. Our value lies in replacing the "middleman" fee with an automated, high-speed matching engine.

What we’re proud of

Identified a massive, underserved market — a $2T infrastructure boom that will need millions of skilled workers.
Built AI that empowers rather than replaces blue‑collar workers.
Created a true win‑win model: companies save $1M+ while workers earn 25% more.
Designed thoughtful AI applications: job parsing, upskilling roadmaps, and smart matching.
Validated everything through real user research (20+ interviews).
Positioned correctly — competing with staffing agencies, not job boards.
Demonstrated concrete financial impact: $10.4M revenue potential with near‑zero marginal costs.

Most importantly: we’re building technology that creates economic opportunity for the people who build America’s infrastructure — and that’s what we’re proudest of.

What we learned

1.⁠ ⁠The Economics are Broken: By eliminating agencies, we found that companies save roughly $832,000 per 100 hires, while workers gain $1.4M in total wages.

2.⁠ ⁠Matching vs. Searching: The bottleneck isn't finding people; it's the 1,000+ manual resume screens. AI solves this instantly.

3.⁠ ⁠Pathway Frustration: Workers want to know which certifications actually increase their paycheck.

What's next for NexSetu

•⁠ ⁠Next 30 Days: Integrate Texas TDLR API for automated license verification.

•⁠ ⁠Months 1-3: Pilot with 3 data center contractors and expand to HVAC/Welding.

•⁠ ⁠Goal: Nationwide expansion across all 50 states within 12 months.

•⁠ ⁠One-Click Compliance: A value-driven feature that automates insurance coverage and IRS paperwork for a nominal fee. We aim to provide the utmost convenience by making legal and administrative tasks a one-click experience.

Built With

alembic
artificial-intelligence
async
axios
bcrypt
fastapi
jwt
lucid
openai
postgresql
pydantic
python
react
router
sqlalchemy
tailwind
typescript
vite

Submitted to

SwampHacks XI

Created by

I originated the project and led the discovery phase, conducting market and user research that validated the opportunity. I architected and implemented a modular, API-first backend framework designed for extensibility and seamless integration by teammates, with clear developer-facing contracts to speed adoption and integrated the AI API key securely (environment variables, secret management and clear developer docs) so teammates can call the models without exposing credentials. Compared our product with key market competitors helping establish and identify unique features needed by our application that will give us a competitive edge. I also created the pitch deck and delivered the project presentation to the judges .

Saaketh Balachendil
My primary focus was building a responsive and modern frontend from scratch. I set up the project with TypeScript and explored Tailwind CSS v4 to build the dynamic Employer and Worker dashboards. I also implemented the real-time profile management system. I learned a lot about rapid prototyping and managing complex state within a short timeframe.

Forrest Yan Sun
My Contribution:

I led the backend development for this AI-powered skilled worker job matching platform. I architected and implemented the complete FastAPI backend infrastructure (2,342+ lines of code), including:

AI-Powered Job Posting System: Built an intelligent job description parser using OpenAI's API that automatically extracts key requirements (certifications, experience, location, pay range) from natural language job descriptions, reducing employer posting time by ~80%.

RESTful API Architecture: Designed and implemented async job management endpoints (CRUD operations) with SQLAlchemy ORM, Pydantic validation, and comprehensive error handling for job creation, updates, retrieval, and deletion.

Database Design: Created normalized database schema with User, Candidate, and Job tables using async SQLite/PostgreSQL support, including proper foreign key relationships and UUID-based identification.

Service Layer Architecture: Developed modular service layers (AI service, job service) separating business logic from API routes, ensuring maintainable and testable code.

Developer Documentation: Authored comprehensive integration guides for frontend developers, including API specifications, example requests/responses, and setup instructions.

I leveraged my industry experience in the field of Applied AI to bring this concept to reality, navigating complex software development challenges to build a resilient system.

Varun Bohara
I’m Varun Bohara, a backend and applied AI engineer with around 3 years of experience, focused on building scalable backend and AI solutions
I contributed extensively to the backend development of the project, focusing on building robust systems for recording and managing candidate information. This involved designing and implementing database models, API endpoints, and validation logic to ensure that candidate profiles are accurately captured and easily retrievable. My work enabled seamless integration between the frontend and backend, allowing employers to view detailed candidate data and workers to manage their profiles efficiently.

In addition to backend engineering, I played a key role in developing the AI-powered job matching feature. I implemented the logic that leverages large language models to match candidates with jobs and vice versa, ensuring that both employers and workers receive intelligent, personalized recommendations. Beyond coding, I also took charge of recording the demo and editing the final video, helping to clearly showcase our platform’s capabilities and user experience to stakeholders and judges.

Akshay Dhawale