Connecting job owners & candidates

Purpose
This project aims to bring structure and clarity to unstructured academic job advertisements. By extracting and standardizing key information from each posting, we can make job posts easier to search, compare, and align with suitable candidate profiles.

How it works
Our tool analyzes job ads line by line. It identifies and classifies the main research field (Data Science, Economics, Neuroscience), sub-fields (Algorithmic Fairness, Behavioral Economics, Neuroimaging), must-have and desirable skills, and the core research mission or objective of the position.

Why it matters
By classifying each job consistently, we can match postings against candidate profiles more objectively. It also helps vacancy holders clarify requirements, so candidates know exactly what the position involves.

Example

• Original job posting: PhD on Algorithmic fairness for data-driven decision making
• Main field: Data science
• Sub-field(s): Machine learning / Ethical AI
• Must-have skills: Programming (Python/R), background in fairness metrics, big data experience
• Nice-to-have skills: Knowledge of optimization, control theory
• Mission: Identify and reduce bias in automated decision-making across various applications (finance, health, public policy, etc.)

Purpose
This project focuses on converting candidate CVs (whether PDF, Word, or even scanned images) into a structured format that aligns with the standardized labels used in job postings.

How it works
Our system first anonymizes the CV by removing all GDPR-sensitive information, such as name, age, gender, and contact details. Candidates are asked to verify this anonymized version before further processing.

Then, the system analyzes each CV and extracts key elements such as: field/sub-field, skills, publications, education, work experience, and career goals. Each element is categorized and standardized, creating a consistent structure across very different-looking documents.

Why it matters
CVs come in all shapes and styles, some have photos, fancy designs, or bullet lists. This variation makes them hard to compare objectively.
By converting them into a standardized format, we can:
- match candidates to academic job postings more effectively
- reduce bias introduced by formatting or visual design
- support HRM professionals in shortlisting with greater confidence
- ensure that aspects like ethnicity, gender, or age play no role in selection.

Example
The original candidate’s CV includes: “PhD candidate in explainable AI with experience in Python, TensorFlow, big data pipelines, and a strong interest in bias mitigation.” The extracted labels might look like:

• Field: Data science
• Sub-Field: Natural language processing, Algorithmic fairness
• Skills: Python, TensorFlow, big data, bias mitigation
• Goal: Advance responsible AI research in an academic or applied R&D setting

Purpose
This project helps candidates complete their academic profiles more easily and meaningfully. The system generates an online form based on a CV, providing candidates with a summary of their background, along with suggestions for possible career paths.

How It Works
The extracted information is turned into a draft profile. Candidates review and edit the suggestions, confirming top skills or adding more details.

Why It Matters
It saves candidates from having to enter the same information repeatedly. It also highlights new directions they may be qualified for, such as transitioning from data analytics into algorithmic fairness, or new academic or applied research directions they may not have considered.

Example
For someone with a strong AI ethics background, the form might propose:

• Current position: PhD candidate
• Future paths: Postdoc in Ethical AI, Industry R&D in Fairness, or an early-career academic track
• Suggested timelines: “Apply for postdoc in 1–2 years” or “Pursue academic roles in 3–5 years if you continue publishing on fairness topics.”

Purpose
This project provides a quick first check to see whether a candidate meets the baseline requirements of a particular job.

How It Works
It compares the job’s required tags (like Python, Machine learning, Fairness metrics) with the candidate’s tags, then calculates an overlap score from 0 to 100.

Why It Matters
It filters out obvious mismatches (e.g., a candidate with no programming experience for a coding-heavy job). This helps recruiters or hiring committees save time, by quickly focusing on relevant candidates.

Example
For a 'Machine Learning Researcher' job requiring Python, ML frameworks, and fairness knowledge, if a CV mentions 'Python, TensorFlow, studied ML ethics', simple matching might yield a 70–80% score, enough to move the candidate to the deeper analysis stage.

Purpose
This project goes beyond concept matching. It identifies deeper connections between candidate profiles and job requirements, uncovering hidden strengths (synonyms, related research areas) and pinpointing specific skill gaps.

How it works
The system interprets CVs and job descriptions contextually. For example, it understands that “YOLOv5” and “Scikit-learn” both relate to advanced machine learning frameworks. It also detects missing elements, such as a job requiring big data experience when a CV only describes work on small datasets. The outcome is a personalized “gap sheet” that highlights areas for growth and suggests relevant resources or training.

Why it matters
Instead of dismissing candidates missing one or two skills, employers can see where a small investment (e.g., a short course) could close the gap. This promotes fairer, more inclusive hiring and allows candidates to grow into roles rather than being excluded too soon.

Example
A candidate might have strong Python and AI ethics skills, but limited hands-on experience with advanced control theory. The system flags that gap and suggests a tailored reading list or crash course, helping both the candidate and the employer make an informed decision.

Purpose
This project displays the best matches in a clear, interactive format for both job owners and candidates.

How it works
Job owner dashboard: Shows a ranked list of candidates with color-coded strengths and gaps. The user can quickly invite candidates or ask for more details with one click.

Candidate dashboard: Provides a similar view of matching jobs, allowing candidates to see which roles best fit their profile so they can apply or request more info.

Why It Matters
It eliminates the need to wade through inboxes or endless job listings. Everything is organized in one place, enabling quick decisions and smoother interactions with simple 'accept', 'invite', or 'skip' actions.

Example
For a PhD role in Algorithmic Fairness, the dashboard might highlight three top-scoring candidates, with one showing an 85% overall match. That candidate’s primary gap, 'limited control theory knowledge', appears in yellow. The job owner can either invite the candidate immediately or propose a short skill assessment.

Purpose
This project provides mini-tests, mock interviews, and real-time feedback, so near-miss candidates can quickly address skill gaps to stay in the running.

How it works
The system generates short quizzes or coding tasks focused on a candidate’s weakest areas. It can also simulate interview questions tailored to a job, providing feedback on the candidate’s answers and suggestions for improvement.

Why it matters
Instead of a simple yes/no decision, candidates get a chance to demonstrate potential and grow. Employers see how quickly applicants adapt, which can be more important than having all the skills on day one.

Example
Potential interview questions for a job requiring fairness interventions in ML:
- “Why is algorithmic fairness crucial in real-world ML systems?”
- “Describe a bias mitigation project you worked on, and the methods used.”
- “In what ways can optimization theory help maintain fairness constraints?”
- “How would you validate the effectiveness of fairness interventions in a large dataset?”
- “Discuss your programming and data-driven modeling experience.”

Candidate’s Answers
The candidate describes working on a credit scoring project where they used reweighting and adversarial debiasing to address income bias, mentioning the use of Python for an end-to-end data pipeline, focusing on different demographic variables.

System’s Feedback
Points out strengths in bias methodology and suggests expanding on validation metrics for stronger responses. This feedback loop helps the candidate refine answers before an official interview.

Job owners save time filtering top candidates and highlighting their strengths and gaps. It also promotes fairness: rather than dismissing someone lacking a single skill, you offer the chance to learn and possibly discover a hidden star. By focusing on both technical and personal qualities, it becomes easier to find people who are a good fit for the role and the organizational culture.

Candidates experience less redundancy. Much of their profile is auto-filled, in stead of re-entering data. They also gain clarity by seeing exactly which jobs align with their profile. Most importantly, they have an opportunity to grow through mini-quizzes, gap sheets, and mock interviews; so a near miss becomes a future success.

Ultimately, everyone benefits. It’s a data-driven approach to academic and research hiring that accelerates finding the best matches in a fair and transparent way. The feedback system also lifts “almost-ready” talent to “fully ready”, helping to close skill gaps in real time.