Omnichannel AI Chatbot Development with Unified Core

Imagine: a client writes in Telegram, clarifies details, attaches a photo. They switch to your website — and the bot doesn't remember the conversation. According to Zendesk Customer Experience Trends Report 2023, 70% of users leave if they have to repeat information. Lost conversions are not a hypothesis but direct losses. Our omnichannel AI chatbot uses a unified AI core to maintain cross-channel context: one trained model, one knowledge base, one analytics. We develop such systems with 5+ years in AI/ML and 50+ delivered projects. Average response time drops by 40%, cost per ticket by 30%. Investment for a basic two-channel setup starts at $50,000. Our bots achieve 2x higher user retention compared to single-channel bots. Get a consultation on your bot's architecture — we'll help select the optimal stack.

Omnichannel Benefits and User Identification

Without omnichannel, each channel is an isolated bot. The client switches and repeats information, gets annoyed, leaves. A unified core provides a single context: a user can start in Telegram, continue in WhatsApp, finish on the website — and the bot remembers the entire conversation. This increases containment rate by 1.5x compared to disjointed bots, and customer LTV by 25%. Support load reduction reaches 50%. Response accuracy is up to 30% higher than traditional chatbots.

The global challenge is user identification across channels. We use strategies: authorization via a single website account (linking messengers), requesting a phone number via Telegram API with subsequent matching to WhatsApp, email verification to create a unified profile. Without explicit linking, profiles are considered different — this is technically honest and protects privacy.

Architecture and Channel Adapters

[Telegram]  [WhatsApp]  [VK]  [Viber]  [Web Widget]  [Instagram]
     ↓           ↓        ↓      ↓           ↓             ↓
[Channel Adapters — message format normalization]
     ↓
[Unified Message Router]
     ↓
[Core AI Engine]
    ├── Intent Recognition
    ├── Context Manager (Redis: user_id → conversation_state)
    ├── RAG / Knowledge Base
    ├── Tool Executor (CRM, ERP, DB)
    └── Response Generator
     ↓
[Channel Adapters — formatting per channel]
     ↓
[Telegram]  [WhatsApp]  [VK]  ...

How to Write an Adapter for WhatsApp? (Step-by-Step)

1. Receive incoming message via WhatsApp Business API (webhook).
2. Parse JSON: extract from, text.body, timestamp, attach media if present.
3. Create a UnifiedMessage object:
   • channel = "whatsapp"
   • user_id = f"whatsapp:{from}"
   • text = body.text
   • media = [MediaItem(url=...)]
4. Pass to the router. Receive a response from the core.
5. Build the reply: if buttons are present, form an interactive object; if text, use text.
6. Send a POST request to /{phone-number-id}/messages.

Such an adapter takes 2–3 days to write for a standard channel.

Message Normalization

UnifiedMessage — a unified format:

from dataclasses import dataclass
from typing import Optional, List
from datetime import datetime

@dataclass
class MediaItem:
    url: str
    mime_type: str

@dataclass
class UnifiedMessage:
    channel: str          # "telegram", "whatsapp", "vk", "web"
    user_id: str          # global ID (channel:original_id)
    text: Optional[str]
    media: Optional[List[MediaItem]]
    timestamp: datetime
    metadata: dict        # channel-specific data

Each channel converts incoming messages to UnifiedMessage and outgoing messages back to the channel-native format.

Dialogue Context and User Linking

Redis stores dialogue state with a TTL of 24 hours:

import redis
import json

class ConversationContextManager:
    def __init__(self):
        self.redis = redis.Redis(decode_responses=True)

    def get_context(self, user_id: str) -> dict:
        data = self.redis.get(f"ctx:{user_id}")
        return json.loads(data) if data else {"history": [], "profile": {}}

    def update_context(self, user_id: str, update: dict):
        ctx = self.get_context(user_id)
        ctx.update(update)
        ctx["history"] = ctx["history"][-20:]
        self.redis.setex(f"ctx:{user_id}", 86400, json.dumps(ctx))

Comparison of Profile Merging Strategies

Analytics, Scaling, and MLOps

An omnichannel platform provides cross-channel analytics: channel mix, cross-channel journey, containment rate per channel, reasons for switching. Stack: ClickHouse for events, Grafana for real-time dashboards. This allows identifying bottlenecks and optimizing routes. Support budget reduces by 40% through automation.

Horizontal scaling of the Core AI Engine (stateless + Redis), Message Queue (Apache Kafka) between adapters and the core, rate limiting, graceful degradation. MLOps processes: model quality monitoring, A/B testing of prompts, automatic retraining on metric drops. This guarantees stability under loads up to 10,000 RPS.

What Is Included

• Architecture documentation: interaction schema, stack selection, solution justification.
• Source code: AI core, channel adapters, integrations with Redis/Kafka.
• Integration with CRM, ERP, knowledge base (RAG).
• Prompt configuration and toxicity monitoring system.
• Client team training (2 days) and 3 months of warranty support.
• Analytics dashboards (Grafana) and alerting.

Implementation Timeline

• Month 1–2: Core AI engine, first channel (Telegram), basic context management.
• Month 3: Add 2–3 channels (WhatsApp, VK, Web), unified analytics.
• Month 4–5: CRM/system integrations, agent handoff, advanced personalization.
• Month 6: Load testing, monitoring, production launch.

To evaluate your scenario, contact us. Get a consultation on architecture and a preliminary budget estimate in 1–2 days. Typical project costs range from $50,000 to $150,000 depending on complexity.

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.