AI Resume Screening System: Automate Candidate Selection with LLMs

Imagine a recruiter spending 6–10 minutes to skim a single resume. With 500 applications per vacancy, that's 50+ hours of continuous reading. We automate this stage: an AI system scans each resume, extracts key data, evaluates fit against job requirements, and outputs a ranked list with justifications. The result — instead of a pile of files, you get a ready candidate pipeline with comments and interview questions.

Resume screening powered by LLMs cuts time-to-screen from hours to minutes and reduces hiring costs by up to 40%. AI screening is 10x faster than manual and drops missed candidates from 30% to 5%.

How AI screening works

Our system accepts resumes in any format — PDF, DOCX, direct input, or via hh.ru API. After parsing, data is normalized: experience recalculated to full months, skills mapped to a unified vocabulary (e.g., "Python + Django" mapped to web-framework knowledge). Then an LLM (GPT-4o, Claude 3.5, or Llama 3) compares the profile against the vacancy text along multiple axes: technical match, experience relevance, education.

class ResumeScreeningResult(BaseModel):
    candidate_name: str
    match_score: float           # 0-1
    technical_match: float       # fit to technical requirements
    experience_match: float      # fit to experience
    education_match: float       # fit to education

    strengths: list[str]         # candidate strengths
    gaps: list[str]              # gaps relative to requirements
    highlight_skills: list[str]  # key skills from resume

    recommendation: Literal["strong_yes", "yes", "maybe", "no"]
    reasoning: str               # 3-5 sentence justification
    suggested_interview_questions: list[str]

def screen_resume(resume_text: str, job_description: str) -> ResumeScreeningResult:
    return llm.parse(
        build_screening_prompt(resume_text, job_description),
        response_format=ResumeScreeningResult
    )

Resume parsing

Resumes arrive in various formats: PDF, DOCX, hh.ru API. Extraction of structured data:

• hh.ru API: resumes are already structured (JSON)
• PDF/DOCX: unstructured.io or custom vision-based parser
• LinkedIn: LinkedIn Talent Solutions API (paid)

Normalization: experience dates → working months, skills → standard dictionary.

Why AI screening beats manual

Model comparison for screening

Model selection depends on privacy and budget. For sensitive data, we use Llama 3 on on-premise GPUs — inference savings can exceed $2,000/month.

Preventing discrimination

AI can replicate discriminatory patterns from historical data. Our mitigations:

• Removal of demographic data before evaluation (name, age, photo)
• Regular audit: no systematic bias by gender, age, or university
• Transparency: every rejection must be justified by professional criteria

Under employment law (e.g., 64-FZ): discrimination based on non-professional traits is prohibited.

ATS integrations

• hh.ru for employers: API for bulk screening of responses
• Huntflow: REST API for pipeline automation
• 1C:Salary and HR: candidate integration into HR system
• Potok.io / Talantix: native integrations via Webhook

Metrics: time-to-screen (from hours to minutes), quality-of-hire (% of AI-recommended hires still in company after 6 months), recruiter satisfaction score.

Development process

1. Analysis — study your current hiring process, resume volume, accuracy requirements.
2. Design — select model (GPT-4o, LLaMA 3, Mistral), define skill vector representation.
3. Implementation — build parsing pipeline, evaluation prompt, result caching system.
4. Testing — calibrate on historical data: accuracy, speed, false positive rate.
5. Deployment — deploy on your infrastructure (AWS, on-premise), connect ATS.
6. Monitoring — track model drift, update prompts.

What's included

• Research of your HR process and specification preparation
• LLM selection and calibration to your screening criteria
• Development of parsers for non-standard resume formats
• Integration with ATS and HR systems (API, Webhook)
• Documentation and training for the recruiting team
• 1 month of warranty support after launch

We have 10+ years of experience in AI/ML, 40+ completed projects, and have been automating recruitment for companies from startups to enterprise for years. Our systems process up to 10,000 resumes per day without quality loss. Contact us for a case evaluation. Get a consultation on implementation — prototype within 2 weeks.

NLP Development: Text Classification, NER, Embeddings, and Information Extraction

We often receive a task: process 50,000 support tickets — currently all manual. Dataset — 3,000 labeled examples, 12 categories, imbalance: one category occupies 40% of the sample, three at 1-2% each. Baseline accuracy — 78%. Sounds decent until you look at recall for rare classes: 0.31, 0.44, 0.28. These classes — complaints and churn threats — are most important to the business.

This is a typical NLP development project. The problem is not the algorithm but that accuracy is the wrong metric. Our experience across 30+ projects shows: we start by analyzing business metrics and only then choose the model.

Why accuracy is not the right metric for rare classes?

Accuracy ignores imbalance. If the "churn" class appears in 2% of cases, the model can predict "all good" and get 98% accuracy — but the business loses clients. Solution: F1 macro (averaged over all classes) or weighted F1. For NER — strict entity F1 (exact matches only). We guarantee: after choosing the correct metric, model quality becomes measurable and predictable.

Text Classification: From BERT to Distillation

BERT-like models are the standard for classification. ruBERT-base or ruBERT-large from DeepPavlov for Russian. multilingual-e5-large — for multiple languages in one pipeline. XLM-RoBERTa-large — a strong multilingual backbone.

Fine-tuning for classification: add a classification head on top of the [CLS] token, train for 3-5 epochs with lr=2e-5, weight decay=0.01. For imbalance — weighted CrossEntropyLoss or focal loss with gamma=2.0. Contact us — we will show a code snippet.

Imbalance case study. Dataset — 3,000 examples, imbalance 1:20. Solution: class_weight via sklearn + CrossEntropyLoss. Additionally — augmentation of rare classes via backtranslation (ru→en→ru through MarianMT). Recall for rare classes rose from 0.31 to 0.67 with a slight drop in accuracy (76%→74%). Full NLP development end-to-end took 3 weeks.

Distillation for production. BERT-large gives F1 0.89, but inference on CPU — 180ms. Distillation into DistilBERT or ruBERT-tiny2 reduces latency to 25ms with F1 0.84. Export to ONNX Runtime provides an additional 1.5-2x speedup. DistilBERT achieves 7x lower latency than BERT-large with only a 5% drop in macro F1 – a typical production trade-off.

How to choose between BERT and LLM for your task?

For most classification and extraction tasks, BERT-sized models offer the best trade-off between cost and performance. Shift to LLMs only when the task demands generation, complex reasoning, or zero-shot generalization.

NER: Named Entity Recognition

NER — extracting persons, organizations, locations, dates, amounts, document numbers. For general categories (PER, ORG, LOC), pre-trained models work well. For specialized ones (medical terms, legal concepts) — fine-tuning is needed.

Data annotation. The main cost of an NER project. For a quality model — 500-2,000 labeled sentences per entity type. Tools: Label Studio (open source) or Prodigy (by spaCy creators). IOB2 format — standard.

Architecture. Token classification on top of BERT: each token gets a label (B-PER, I-PER, O). spaCy 3.x with transformer pipeline — a convenient production choice.

Nested entities. Standard IOB models cannot handle nested entities (organization inside an address). For such tasks — span-based NER: SpanBERT or SpERT. More complex but correct.

Post-processing is mandatory. The model predicts tokens — normalized entities are needed. Date — dateparser. Amounts — regex + validation. Names — deduplication via rapidfuzz. Included in our standard delivery.

Sentiment Analysis and Opinion Mining

Binary classification positive/negative works out of the box with BERT. Complexity — aspect-based sentiment analysis (ABSA): "the restaurant has good food but terrible service." For ABSA: aspect extraction (NER) + sentiment per aspect. Joint models BERT-for-ABSA — quality on Russian data is lower due to dataset scarcity. RuSentiment, SentiRuEval — main resources.

For production with simple positive/negative/neutral: distil models are enough. Three classes, balanced dataset, 2,000+ examples — F1 macro 0.82-0.87 in 1-2 days.

Text Summarization

Extractive summarization (select sentences) — TextRank or BM25 without training. Fast, no hallucinations. Good for long documents.

Abstractive (generates new text) — seq2seq: mT5, mBART, FRED-T5, ruT5-large. For production via LLM API (GPT-4, Claude) — often the best cost/quality/speed trade-off.

Embeddings: Vector Representations of Text

Embeddings are the foundation of semantic search, deduplication, clustering, RAG. Quality critically affects downstream tasks.

Models. E5-large-v2, BGE-M3, multilingual-e5-large — strong multilingual embedders. sentence-transformers/paraphrase-multilingual-mpnet-base-v2 — fast option. For Russian: ru-en-RoSBERTa (Skoltech) performs well on semantic textual similarity.

Embedding quality evaluation uses the MTEB benchmark as standard. But top results on MTEB don't guarantee success on a domain dataset — we build domain-specific eval.

Fine-tuning embeddings. If standard models don't give the required Recall@k — contrastive learning on domain pairs with MultipleNegativesRankingLoss. How to perform this for domain data:

1. Collect 500–2,000 semantically similar pairs from your domain.
2. Apply MultipleNegativesRankingLoss with a batch size of 32–64.
3. Train for 1–3 epochs using AdamW (lr=2e-5).
4. Evaluate Recall@k on a held-out domain test set.

This approach yields a 5–15% improvement in Recall@k in practice.

Dimensionality and storage. E5-large: 1024 dim, float32 — 4KB per vector. For 10M documents — 40GB. Quantization int8 reduces to 10GB. FAISS IVF_PQ — more compact but with losses. Included in our deployment recommendations.

Information Extraction

Structured extraction is a frequent task. Examples: key contract terms, technical characteristics, dates and amounts from invoices.

1. Regex + rule-based. For INN, OGRN, amounts, dates — more reliable than neural networks. No data required.
2. NER + post-processing. For variable formats.
3. LLM with structured output. GPT‑4 / Claude with JSON schema — for complex documents. Cost: minimal per document. For 10k+ documents/day — we calculate the economics.

We guarantee a hybrid: regex/NER for typical fields + LLM for edge cases. Our guarantee is backed by years of production experience and more than 30 projects.

Work Stages

Prototype on existing data — 1-3 weeks. Production system with CI/CD — 1.5–2.5 months. Cost is calculated individually — get a consultation for a project estimate.

What's Included

• Model and pipeline architecture documentation
• Access to the model via REST API (FastAPI + ONNX)
• Client team training (2-hour webinar + Q&A)
• Accuracy guarantee on the agreed test set
• Months of post-delivery support (bug fixes, adaptation to new data)

Our Experience

Years of NLP projects from classification to RAG systems. The team includes ML engineers experienced with Hugging Face, spaCy, LangChain, MLOps. We use vLLM, Kubeflow, Weights & Biases — a production stack, not toys. Contact us to evaluate your NLP project within two days — request a free consultation on your text processing pipeline.