Data cannot leave your perimeter
If security policy or subsoil-use requirements prohibit sending data to external clouds, our default architecture — on-premise or in the customer's private cloud — fits without compromise.
Industrial AI · Kazakhstan
Industrial AI for Kazakhstan's plants: we see the failure before the breakdown
Predictive maintenance, computer vision for safety, digital twins and private LLMs for oil & gas, uranium, power and mining. Deployed on your servers — data never leaves your perimeter. A pilot in 8–12 weeks with a measured effect.
- 8–12
- weeks to a working pilot
- 20–40%
- reduction in unplanned downtime
- 100%
- of data stays within your perimeter
Astana Hub residentinternational technopark, LLP under Kazakhstan law- On-premise by defaultdata never leaves your perimeter
- NDA from the first contactconfidentiality before any details are discussed
- Pilot in 8–12 weekssuccess criteria agreed before the start
How it works
Plant data becomes forecasts and recommendations without leaving your perimeter
We connect to what you already have: SCADA, CMMS, cameras and documentation. Models run on your servers, and the result is measured in your KPIs — downtime, availability, specific consumption.
What we connect
- SCADA / DCSOPC UA, Modbus, PI System
- CMMS / EAMSAP PM, 1C, maintenance history
- CamerasRTSP streams from cameras you already have
- Documentationstandards, procedures, LIMS data
Enterprise perimeter · on-premise
105 ML platform
on your servers or in a private cloud inside the perimeter
- Failure forecast and remaining life
- Event detection on video
- Process optimization
- RAG assistant over documents
Zero outbound traffic · read-only access to control systems · access logging
What you get
- Condition-based maintenancefailure warning days to weeks ahead
- HSE incidents 24/7PPE, danger zones, hot work
- Operator recommendationssetpoints from the best shifts, explained
- Effect in KPIsdowntime, availability, consumption, EBITDA
Solutions
Four classes of problems we solve
We turn SCADA, process-control, camera and maintenance data into working models. Every solution runs inside your perimeter and is measured in your KPIs.
01
Predictive equipment maintenance
Models on vibration, SCADA telemetry and maintenance history predict failures of pumps, compressors, turbines and gas-pumping units days to weeks ahead — repairs are planned before the breakdown.
20–40%fewer unplanned outages
02
Computer vision for HSE
Video analytics on the cameras you already have: PPE compliance, people in danger zones, hot work, leaks and smoke. Edge inference — video never leaves the site.
24/7automatic incident logging
03
Digital twins and process optimization
Models of units and process stages: regime forecasting, setpoint advisor based on the best shifts, energy and reagent optimization — with no writes to the control system.
3–8%process efficiency gain
04
Private LLMs inside your perimeter
Local language models on your infrastructure: answers from standards and procedures in seconds with a citation down to the clause, assistants for engineers and dispatchers, reporting automation.
On-premisenot a single token leaves the perimeter
Our own product · live demo
Boljam: failures visible 13 days before vibration monitoring flags them
An RMS velocity threshold is a reactive tool: by the time a machine reaches zone C the bearing is already failing. Boljam is a working predictive-maintenance system that catches the defect in the envelope spectrum long before the threshold. Open it and check for yourself: vibration diagnostics, telemetry, maintenance history, 3D twins and economics in one loop.
- ISO 20816-3
- Envelope spectrum
- Remaining life P10/P50/P90
- 3D digital twins
- Rod-pump dynamometer cards
- Maintenance economics
boljam.105.kzlive
- days of median lead time over the RMS thresholdacross 377 machines in the demo fleet
- 13
- pieces of equipment under monitoring1,302 measurement points, 147 MW
- 450
- ROC-AUC of the anomaly detectorfalse-alarm rate 0.001
- 0.977
- months of history — scrub with the cursorsignal recomputed in the browser
- 18
A bearing 85% destroyed — instruments still read “as new”
A pump at 1476 rpm with an 85% outer-race defect: RMS velocity of 1.86 mm/s — zone A under ISO 20816-3. Formally the machine is healthy. Meanwhile the envelope at the defect frequency has grown 95-fold and kurtosis has gone from 2.96 to 10.87. That is exactly what Boljam surfaces.
Demonstration on synthetic data: the figures come from a physics model, not from measurements on real equipment. Asset names are taken from public sources; the system is not the reporting of any company and uses no corporate branding.
Process
From data audit to industrial deployment in three steps
Success criteria and the method for measuring the effect are agreed before the start. Every stage ends with a result you can show the board.
- 2–3 weeks
Data & process audit
On-site visit, readiness assessment of SCADA / process-control / CMMS data, and a retrospective analysis on an anonymised export where needed.
Stage result
Business case with effect estimate and a pilot plan
- 8–12 weeks
Pilot on a single site
A working model on the customer's real data; the forecast is checked against actual failures and work orders from a past period.
Stage result
Model, quality metrics, economic effect report
- per pilot plan
Industrial deployment
Integration with your systems (SCADA, SAP, 1C, LIMS), staff training, handover of rights and SLA support.
Stage result
System in operation, trained team, SLA
NDA from the first contact. Ready for national company and JV procurement procedures.
Industries
Built for Kazakhstan's industrial sector
We understand production, process-control data and the security requirements of subsoil users — and speak the language of chief engineers: availability, specific consumption, permits to work.
Oil & Gas
ESP, pump and compressor failure prediction, production and waterflood optimization, HSE video monitoring at fields and refineries.
- ESP
- SCADA
Uranium & Mining
In-situ recovery optimization, predictive maintenance for mining equipment, core analysis, beneficiation and hydrometallurgy.
- ISR
- CMMS
- processing
Energy
Load and imbalance forecasting, predictive maintenance for generation assets, transformer DGA, grid digitalization.
- generation
- grid
- DGA
Manufacturing
Digital twins of process units, process quality control, LLM assistants for technical documentation.
- DCS
- LIMS
Why now
2026 is Kazakhstan's Year of Digitalisation and AI: national companies have KPIs and budgets
Figures come from public sources. A local team with on-premise architecture closes this KPI faster and 2–3× cheaper than foreign integrators.
- 62
- AI projects under way across the Samruk-Kazyna group
- $1.3 bln
- expected effect of the group's AI projects by 2030
- +5%
- EBITDA growth KPI driven by AI for portfolio companies
- 12K t
- of additional oil in 2025 — ABAI AI waterflooding module (KMG)
sk.kz, 2026
sk.kz, 2026
Samruk-Kazyna
KazMunayGas
Why 105
When we are a good fit
105 Industrial AI ("105kz" LLP) is a Kazakhstan-based developer of industrial AI systems and an Astana Hub resident. We are not a general-purpose integrator: we specialize in heavy industry and air-gapped environments. These are the scenarios where we are strongest.
Honest about our limits
Boxed cloud SaaS, ERP implementations and custom web development are not our profile. If your task is in that area, we will say so at the first meeting and point you in the right direction where we can.
Deployment on site in Kazakhstan
A local team travels to your facilities, connects to process-control and SCADA systems on site and supports the system within Kazakhstan's jurisdiction — no time zones or visas.
Russian and Kazakh language support
Interfaces, reports, staff training and LLM assistants in the languages your plant actually works in. Contracts and support under Kazakhstan law; English for international JVs.
Heavy-industry domain focus
Oil & gas, uranium (ISR), power and mining: we work with SCADA telemetry, vibration diagnostics, DGA and maintenance histories rather than adapting generic office products.
Astana Hub resident
Technopark tax incentives, transparent company details and a Kazakhstani LLP — a counterparty your security department does not need explained.
Procurement-ready
Samruk-Kazyna group and JV procurement procedures, domestic producers register, GOST 34/19, staged acceptance and an NDA from the first contact.
Company details"105kz" LLP · BIN 251140032193 · 42/16 Timiryazev St., Almaty, 050057, Kazakhstan
About the companyFAQ
Frequently asked questions
What chief engineers, security and procurement teams usually ask — with straight answers.
Next step
Discuss a pilot project
Tell us about your task — we'll come back with a data audit plan and an effect estimate within two business days.
- 1A 30-minute call: the task, the data, who signs off on security
- 2Data audit or a retrospective analysis on an anonymised export
- 3An 8–12 week pilot with success criteria agreed before the start
NDA from the first contact
Reply within one business day