SignalForge

SignalForge is a resume market intelligence studio.

Instead of acting like a slightly nicer listings interface, it treats the resume as the starting point and turns it into three things:

• a live market read,
• an opportunity map,
• and a next-move strategy called Signal Paths.

The core idea is simple: most people do not need more raw listings. They need a clearer read on where their resume lands now, where it can stretch next, and what signals are showing up repeatedly across real opportunities.

That is what SignalForge is built to do.

What SignalForge Produces

When someone uploads a .pdf, .docx, or .txt resume, the app:

1. reads the resume text,
2. extracts the strongest skills, role signals, and phrases,
3. searches live U.S. opportunities from structured company-board feeds first,
4. ranks the strongest matches with local semantic scoring,
5. turns the results into a market-facing dashboard instead of a plain list.

The output is not just "here are some openings."

It is:

• Resume Market Read: how the uploaded resume is being interpreted across the strongest results,
• Opportunity Map: a ranked view of openings that fit the current profile,
• Signal Paths: three strategic directions based on the shortlist.

Signal Paths

This is the feature that makes the product feel different.

Signal Paths converts the shortlist into three next moves:

• Ready Now: where the resume is already landing strongest,
• Stretch Next: the closest adjacent lane worth exploring,
• Unlock More: the recurring asks that could widen the search.

That makes the product feel more like a positioning and navigation tool than a standard recommendation engine.

This is already live in the shipped product. Signal Paths appears on the results page as three strategy cards:

• Ready Now
• Stretch Next
• Unlock More

Why This Feels Different

A lot of products in this space stop at one of these:

• keyword search,
• resume parsing,
• generic recommendation feeds,
• fake certainty scores dressed up as AI.

SignalForge is meant to sit in a more interesting place:

• resume-first instead of search-box-first,
• strategy-oriented instead of list-oriented,
• local-AI-powered without asking users for their own API key,
• honest about fit estimates instead of pretending to predict outcomes.

The core experience runs on bundled local embeddings and heuristic scoring. There is an optional remote-enrichment path in the codebase for experimentation, but the shipped product does not depend on OpenAI or on users bringing their own key.

What The App Actually Uses

Under the hood, SignalForge blends:

• local open-source embeddings,
• TF-IDF overlap,
• title and role alignment,
• role-family consistency,
• recurring skill and phrase signals,
• U.S. location filtering,
• remote / hybrid / onsite mode filtering.

The result is a heuristic ranking layer, not a hiring oracle.

What The Scores Mean

The percentages are fit estimates used to prioritize what to open first.

They are grouped into:

• Strong fit
• Solid fit
• Possible fit
• Stretch

They are intentionally not framed as probabilities or promises.

What Is Already Live

SignalForge already ships with:

• a Resume Market Read that summarizes how the uploaded resume is landing across the current shortlist,
• Signal Paths on the results page,
• matched-signal chips and gap chips on every result card,
• a visible scoring breakdown for each opening,
• live filters for lane, source, fit band, mode, and search,
• a top-five quick-open flow for the strongest opening set.

This matters because the project is not just describing a product direction. The core workflow is already built and deployed.

Why The Ranking Is Trustworthy

The ranking is not meant to be a black box.

There are three trust layers built into the product:

1. Relative fit, not fake certainty A score means the opening ranked strongly inside the current live pool retrieved for that resume. It does not pretend to predict hiring outcomes.
2. Visible evidence on each card Every opening shows matched signals, common gaps, a reason line, and a scoring breakdown so users can audit why it surfaced.
3. Multiple ranking inputs SignalForge blends semantic similarity, title alignment, role-lane consistency, recurring skill overlap, and source quality instead of relying on one keyword count.

The goal is not to hide the model. The goal is to make the shortlist inspectable.

Current Scope

Right now SignalForge is focused on:

• U.S. opportunities only,
• remote, hybrid, and onsite roles,
• resume-first market navigation for a broad range of professional backgrounds,
• people who want to move from uploaded resume to useful direction quickly.

Honest Limitations

This project is much better than starting from a raw listings feed, but it still has real limits:

• live source quality still matters,
• some companies publish duplicate or near-duplicate openings,
• the system is stronger at ranking and positioning than at understanding very nuanced career pivots,
• it does not auto-apply,
• it does not claim that every result is perfect.

The value is that the shortlist and the strategy layer should be meaningfully better than starting from scratch.

Stack

• Python
• Flask
• scikit-learn
• FastEmbed
• Requests
• BeautifulSoup
• pypdf
• python-docx
• Vercel

Running It Locally

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements.txt
python app.py

Then open:

http://127.0.0.1:5000

Deployment Notes

SignalForge is set up to run as a lightweight Python app on Vercel.

Important implementation details:

• uploaded resumes are processed as temporary files,
• the stylesheet is served from public/styles.css,
• the app does not rely on persistent local storage,
• the semantic model is bundled so end users do not need to configure their own AI provider.

The Real Product Idea

SignalForge is not trying to be another search layer for openings.

It is trying to become a resume-aware market navigation layer:

• what looks strongest now,
• what adjacent lane is closest,
• what signals unlock more range,
• and which live openings are worth opening first.

That is the category-defining idea behind the project.