AI-Based Construction Cost Estimation from Documentation

AI-Based Construction Cost Estimation from Documentation

Estimating documentation in construction involves hundreds of volumes of PDFs, DWGs, and IFCs. For instance, a developer receives 50 volumes of project documentation for a 15-story residential building. A manual estimate takes two quantity surveyors a month and costs a significant amount. Our AI system handles it in 3 days with 95% accuracy — processing 10 times faster. The deviation from a manual estimate does not exceed 2–5% for standard projects; for complex objects, accuracy is 90% with full traceability of each item. Order a preliminary analysis of your documentation — we'll show how much time and money AI can save. According to the Ministry of Construction, the average deviation of manual estimates is 5–7% — AI cuts this figure in half.

Incoming Documentation and Formats

A construction project includes design and working documentation, bills of quantities, and technical specifications. The system accepts PDF, DWG, IFC, Excel, and scanned documents via OCR. A BIM model is not required but increases accuracy by 10–15%.

AI System Architecture

[Upload documentation (PDF, DWG, Excel)]
    → [Section classification]
    → [Bill of quantities extraction]
    → [Equipment specification recognition]
    → [Apply unit rates (FSNS, GESN, TER)]
    → [Calculate direct costs]
    → [Apply coefficients (ISR, OZP, overhead)]
    → [Summary estimate]
    → [Human verification]

How AI Extracts Work Volumes

class WorkItem(BaseModel):
    section: str
    description: str
    unit: str
    quantity: float
    normative_code: str | None
    confidence: float

def extract_work_volumes(document_text: str) -> list[WorkItem]:
    tables = extract_tables(document_text)
    items = []
    for table in tables:
        if is_work_volume_table(table):
            parsed = parse_work_volume_table(table)
            items.extend(parsed)
    text_items = llm_extract_work_items(document_text)
    return items + text_items

Deterministic parsing (pdfplumber, Camelot) yields 99% accuracy for tables. Unstructured descriptions are processed by LLM with few-shot templates. Combining methods ensures reliability.

Comparison of Volume Extraction Methods

Why Does AI Estimate Accuracy Vary?

Accuracy of an AI estimate depends on documentation completeness. With a full set of bills of quantities, deviation from a manual estimate is ±10–15% (acceptable for preliminary evaluation). For tender documentation — ±5–8% with thorough verification. On standard objects after fine-tuning — up to 2%. The more training data, the higher the accuracy. We guarantee results on your projects.

How AI Handles Different Documentation Types

Regulatory Framework and Specifications

Applying Regulatory Data

FSNS (Federal Estimate Normative Base) and regional TER collections contain unit rates for all types of construction work. The system maps work descriptions to GESN/FSNS codes via semantic search (1536-dim vector embeddings), obtains normative indicators, and multiplies by current conversion indices. Complexity: work descriptions do not always exactly match FSNS codes — normalization and expert confirmation are needed.

Equipment Recognition from Specifications

Equipment specifications (pumps, boilers, ventilation) are extracted by LLM with linkage to supplier price lists. LLM extracts name, brand, characteristics, quantity. Then it queries current prices via supplier APIs. This is especially relevant for import substitution, where prices change monthly.

Verification and Audit

The system does not replace the quantity surveyor for complex objects — it accelerates their work 5 times. Each estimate section is accompanied by the data source, applied rates, and calculation formulas. Full traceability for audit. Built-in checklists automatically verify mapping correctness and warn about discrepancies. After processing, AI generates a report with all sources and calculation coefficients. The surveyor can check any item and make corrections. The system remembers edits and improves the model for future projects.

Development Process and Timelines

Development includes documentation analysis, dataset collection (at least 5000 pages), model training (fine-tuning LayoutLM + LLM), integration with regulatory database, creation of REST API and modules for estimation software, documentation and training for surveyors, and 6 months of technical support. Each stage includes demonstration of intermediate results.

MVP for a standard residential building — 4–6 months; industrial object with integration — 8–12 months. Contact us for a free preliminary assessment of your project — we'll analyze documentation and calculate timelines.

Our team has 7+ years of experience in AI/ML and has delivered over 30 projects for the construction industry. We work full cycle: from data collection to production. Get a consultation — we'll evaluate your project and propose the optimal solution.

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.