Language models and agents
Fine-tuning LLMs for the client's domain, RAG over corporate data, agentic and multi-agent systems with access to AI tools
We build AIAI for oil & gas
Custom end-to-end development of AI and software systems for companies in Russia and worldwide.
100+
projects for Russian and international clients
~80
engineers in Moscow and Barnaul
R&D center in Barnaul
4
R&D areas: text, voice, face, vision, LLM
6
own products
LLMCore and TeamDisk
SberbankingYandexITRostelecomtelecomSurgutneftegasoil productionNestléFMCGTeboilpetroleum products
ESTESIS develops AI systems and software systems end-to-end for companies in Russia and worldwide. About 80 researchers and engineers in Moscow and Barnaul
R&D team
Research and development are done in-house
We work with modern models: large language and multimodal models, agentic systems, speech and vision foundation models. We adapt them to the client's domain and requirements — fine-tune, train from scratch, distill, test for robustness and put them into production
Voice and speech
Speech models for recognition and synthesis, speaker verification, detection of synthesized and cloned speech.
Biometrics
Face recognition, voice verification, liveness detection, protection against deepfakes and spoofing attacks
Vision and multimodality
Vision-language models, detection and segmentation built on foundation models, training on synthetic and real data
For all areas
Quality assessment
Our own test suites and metrics for the client's task
Red Teaming
Vulnerability testing of models and agents before launch
Distillation and quantization
Compact models for your own servers and edge devices (IoT)
Guardrails
Request and response control, protection of confidential data
Russian and international clients
SberYandexRostelecomSurgutneftegasNestléhuupepathrurbanchain
Industries
AI system architecture
One team is responsible for every block — from data sources to support
Employees
and clients
Cameras
and sensors
Documents
and regulations
Telephony
and recordings
External
systems
Channels
Web
Mobile
Assistants
Kiosks
Business systems
ERP
MES
Industrial control systems
Document management system
AI models
- Computer vision
- LLM agents and RAG
- ML forecasts
- Voice and faces
Client environment
On-prem
GPU cluster
Security and access
Analytics
BI
Reporting
Forecasts
Pipelines
data
Vector
database
Labeling
and datasets
MLOps
and model versions
Monitoring
quality
Operation and support
Portfolio
Team expertise
All cases
Flagship case · Energy and utilities
Heat supply dispatch platform
A domestic platform for the middle and upper levels of the automated dispatch system replacing the previous one: data collection from meters and control of regulators, monitoring, alerts and control of heat substation modes.
Read the case
Transport and IoT · Karusel
Kicksharing platform
Thousands of scooters send data, receive commands and serve rides at the same time. We built the whole platform — hardware, embedded software, cloud and app — as one team.
Karusel
Logistics · Port logistics company
Digital twin of a seaport
A live model of the terminal: cameras, accounting systems and a forecast of berth, equipment and vessel load in one dispatcher window.
Port logistics company
Oil and fuel · Network of 500+ filling stations
Digital ecosystem of a fuel holding
BSP, a loyalty app, a data bus and smart analytics on a data warehouse for a network of more than 500 filling stations.
Network of 500+ filling stations
All projects in the case catalog
Filter by industry and area; the task, solution and result for every project
Client reviews
Quotes are published with the clients' consent
"The task was non-standard; we found no ready-made solutions on the market. The team came from the research side — they showed that a similar problem is solved with methods from an adjacent field, and the prototype confirmed it."
"We started with one pilot four years ago; now we have three systems in production and a shared support setup. In that time we have had two IT directors, and the contractor's team has stayed the same."
Products
We build infrastructure for corporate AI
We build infrastructure
for corporate AI
ESTESIS product line
LLMBox
AI agent platform
An agent builder with tools, a knowledge base and support for any models. Runs in the client's environment or in the ESTESIS cloud.
LLMGuard
AI firewall for LLMs and AI systems
Checks model requests and responses and prevents leaks of confidential data.
LLMGateway
Gateway to LLM models
A single API to 300+ models, routing, usage tracking and cost control.
LLMIndex
Brand visibility in AI answers
Asks AI assistants about your market every day and shows who they recommend instead of you.
LLMRed
Security testing
Tests LLM systems for vulnerabilities and resistance to attacks before launch.
TeamDisk
A corporate storage with an online office and AI agents
Formats
of cooperation
The format is chosen at the first meeting based on the task, timeline and project management model
We consult
You come to us as consultants. We share experience and guide you and your team. We improve your approaches
IT outsourcing
We form a team with our management. A stable team for your product. You set the priorities; we are responsible for the tasks and the team
Outstaffing
Engineers with the right qualifications in your team and by your processes. We handle selection and replacement
Consulting and training
We analyze processes and train the team on your own tasks so that the expertise stays with you.
Support and evolution
Monitoring, incident analysis, model retraining and feature development under an SLA.
Working principles
The stage result is agreed in advance
Before each stage we fix what the client receives at the end and by which metric the result is evaluated
Code and models are handed over to the client
We deploy IT solutions on the client's servers or in a private cloud
Support after launch
The system moves into support: monitoring, AI model retraining, evolution 24/7
Team
Engineers and researchers who design, build and support clients' software systems
Moscow and Barnaul
R&D center in Barnaul and an office in Moscow. We join complex IT projects and see them through to results
Team and careersFAQ
Audit and pilot — a fixed price and a fixed stage result. Development, industrial deployment and evolution are more often done as a dedicated team with a clear rate
It depends on the task, the data and the environment requirements. We start with an audit: based on its results we fix the pilot scope, timeline, cost and success metric. We send the estimate within 5 working days
Yes. We deploy solutions on the client's servers, in a private cloud or in a closed environment. Our own products — LLMBox, LLMGuard, TeamDisk — also run on-premise
Code, trained models and documentation are handed over to the client together with the result of the stage; the scope of rights is fixed in the contract. If the solution uses our products, they are connected under a license
Not only AI. We run the full software development cycle: web and mobile apps, cloud platforms, ERP and MES integrations, embedded software. You can hire a dedicated team or outstaffed engineers
Accredited IT company · Russian Ministry of Digital Development registry No. 20142