AI Transparency

TL;DR

TalentyGo uses artificial intelligence to aggregate job listings, match them to your profile, and generate personalized deliverables (Cover Letter, Mock Interview, Salary Brief, Career Path, Resume Optimization).

We do NOT use protected demographic data (age, race, gender, sexual orientation, religion, disability, national origin) to filter, rank, or discriminate. Matching is based on declared skills, location and sector preferences, and seniority.

1. AI systems we use

OpenAI text-embedding-3-small (1536 dimensions) for semantic embeddings of your resume and job listings. Used to compute cosine similarity between resume and job posting vectors.

OpenAI GPT-4 / Claude for personalized deliverable generation (Cover Letter, Mock Interview Q&A, Salary Brief narrative, Career Path roadmap, Resume Optimization suggestions).

Proprietary classifier v2 for sector categorization (22 sectors: Beauty, Fashion, Tech, Healthcare, Food, etc.). Keyword matching + rule-based scoring; no supervised ML.

2. What we DON'T do

We do NOT screen candidates for employers. We do not offer recruitment or filtering services to third parties. The platform is user-facing: we help the candidate find opportunities, not the employer screen out candidates. For regulatory clarity: this means TalentyGo does NOT fall under NYC AEDT Law scope or EU AI Act high-risk category at this time.

We do NOT use protected demographic data. Matching ignores:

• Age (even if inferable from resume)
• Race, ethnicity, national origin
• Gender, gender identity, sexual orientation
• Religion, political beliefs
• Disability, health conditions
• Marital status, pregnancy

We do NOT retain resume embeddings beyond what's needed. When you delete your resume from your account, embeddings are deleted within 30 days via cleanup cron.

We do NOT sell your data. Ever. See the Privacy Policy.

3. Your rights

Under GDPR (Art. 13-22), CCPA/CPRA, and EU AI Act (Art. 13), you have the right to:

• Know how it works (this page + section 5 below).
• Request explanation of a specific recommendation or deliverable. Email [email protected] with details. We respond within 30 days.
• Opt out of AI matching using traditional filters (city, sector, keyword). Semantic matching is optional and can be disabled on request.
• Request human review of any automated decision affecting you (GDPR Art. 22). Email [email protected].
• Delete your resume and all related embeddings at any time from the Account section.
• California residents (CPRA): right to know, right to delete, right to correct, right to limit sensitive PI use, right to opt out of automated decision-making technology (ADMT).

4. Technical limits and disclaimer

Like all modern AI systems, our models may produce:

• Inaccuracies (e.g. Cover Letter with wrong dates, Mock Interview questions not perfectly relevant).
• Residual bias from the foundation model training data (OpenAI/Anthropic), which we do not directly control.
• False positives in matching (recommended jobs you don't care about) and false negatives (interesting jobs not recommended).

For these reasons, always review AI deliverables before using them with recruiters. TalentyGo does not guarantee specific outcomes from applications, interviews, or hiring. See Terms of Service section 7 (Limitation of Liability).

5. How matching works (optional technical detail)

For the curious: matching between resume and job listing works as follows:

1. Your resume text (extracted from uploaded file) is sent to OpenAI text-embedding-3-small which returns a 1536-dimensional numerical vector.
2. The same process runs for every job listing in our catalog (~30,000+).
3. We compute cosine similarity between your resume vector and each job vector. Higher similarity (max 1.0) means more semantic similarity.
4. We filter by your explicit preferences (cities, sectors, seniority) — here NO AI involved, just traditional SQL filtering.
5. We rank by cosine similarity and show top 50 in your Roster.

Important: we do not aggregate demographic data into embedding vectors. While the OpenAI model might pick up signals from resume text (e.g. "Mario, 35, Italian"), the final ranking depends only on cosine similarity, not on any explicit demographic attribute.

6. Audit and improvement

We're working on:

• Annual bias tests to verify the system doesn't systematically disadvantage protected groups.
• Publishing an AI Impact Assessment once we have sufficient production data.
• Adopting ISO/IEC 42001 (AI Management System) standards when resources permit.

Researchers, journalists, or regulators wanting access to our matching logs for independent audit: contact [email protected].

7. Contact

For questions about AI use at TalentyGo:

📧 [email protected] (GDPR/AI Act requests)

📧 [email protected] (legal matters)