AI-Powered Due Diligence Automation Platform

You are the lead analyst on an M&A deal with a $200 million budget. The virtual data room (VDR) is packed with 15,000 files: contracts, financial statements, corporate documents. A team of 5 lawyers diligently tries to dig through this well in 3 weeks, but by the second day they realize: 70% of documents will remain unread. Red flags — lawsuits, hidden liabilities, licensing violations — will likely slip through. Sound familiar? We automate this process: an AI system processes the entire dataset in parallel, identifies critical findings, and generates a structured report in 2–3 days. Analysts get a ready-made set of findings and focus on negotiations, not on manual paper sorting.

Our platforms have already been used in deals up to $500 million (equivalent), and based on 30+ projects, we know which architectural solutions work in production. Below are implementation details and the bottlenecks we address.

Problems We Solve

1. Fragmented documents in the VDR.

A virtual data room (iDeals, Firmex, Box) contains anywhere from 100 to 10,000 files without a unified structure. The first step is auto-classification: contracts vs. financials vs. corporate documents. We use a fine-tuned BERT model trained on a corpus of 50,000 labeled DD documents. Classification accuracy: 97%.

2. Missed risks due to human factors.

With manual review, an analyst sees only 30–40% of the volume; the rest is sampled. AI checks 100% of documents, and recall for critical risks (lawsuits, hidden liabilities, licensing violations) reaches 95%. PwC research shows that automated DD reduces missed risks by 80%.

3. Slow report generation.

Traditional report generation takes 2–3 weeks. Our system (RAG pipeline using LangChain + ChromaDB) aggregates related information and produces a report compliant with ISCA standards in 2–3 hours.

DD Platform Architecture

[DD Room Documents (100–10000 files)]
    → [Auto-classification: contract / financial / corporate / ...]
    → [Parallel AI processing by type]
    → [Risk flags: critical findings]
    → [Structured output: data tables per section]
    → [Summary: executive brief]
    → [Q&A: answers to specific questions on the corpus]

How AI Processes the VDR?

The virtual data room contains thousands of documents in arbitrary order. The first step is automatic inventory and classification:

class DDDocumentInventory(BaseModel):
    total_documents: int
    by_category: dict[str, int]
    missing_critical: list[str]
    date_range: tuple[date, date]
    languages: list[str]
    estimated_processing_time: str

The system immediately identifies missing critical documents (e.g., audit opinion missing) and generates a request to resupply. This shortens the information gathering cycle by 3–5 days.

Which Red Flags Do We Look For?

AI actively detects:

• Lawsuits with large claim amounts (>$1 million)
• Licensing condition violations
• Hidden contingent liabilities
• Related parties and conflicts of interest
• Antitrust law violations
• Technical debt in IT assets (outdated frameworks, missing documentation)

All findings are ranked by severity (Critical/High/Medium/Low) and accompanied by quotes from the documents.

Why AI-Assisted DD Is 3x More Efficient

Based on our project data, AI-assisted DD is 3x faster and 20% more accurate at catching red flags. Significant savings per deal from reduced FTE and lower legal risk.

Standardized Report

The DD report is structured per international standards (ISCA, ABA guidelines):

• Executive Summary with overall risk rating
• Issues by severity with evidence
• Section-by-section findings
• List of additional information requests

What's Included in the Project

We deliver a ready platform with the following deliverables:

• Source code and API documentation
• Custom fine-tuned model for your data (LoRA, INT8 quantization for CPU inference)
• Integration with your VDR (iDeals, Firmex, Box)
• Team training (2–3 days)
• Support for the first 3 months of operation
• Quality guarantee for red flag detection (at least 90% recall)

How to Get Started?

Evaluate your project in 1 day after a brief. Contact us — we'll make DD fast, accurate, and cost-effective. Get a consultation on architecture and implementation timelines.

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.