PROPRIETARY AI-DCIM ECOSYSTEM | ASTANA, KAZAKHSTAN (UTC+5)

Intelligent 24/7 Remote Monitoring & Predictive Analytics for Modular and Small Data Centers

We replace expensive round-the-clock on-site shifts at container and edge GPU facilities (10 to 50 racks) with automated telemetry harvesting, ML-based overheating forecasting, and remote engineering oversight powered by our proprietary TES Remote DCIM Platform.

25telemetry metrics (SI units)
15 sedge sampling interval
72 hoffline local spool
0inbound ports opened
SITE-EU-01MODULAR HALL A · 12 RACKS · LIVE DEMO

Rack Heat Map · GPU avg °Cclick a rack

COLD AISLE
CRAC-A CRAC-B DRYCOOLER-1
55°90°C

Composite indicesR07

GPU Health Index 94 /100
GPU HEALTH INDEX
PUE 1.28 ISO 30134-2
PUE
RISK SCORE · 1H0.08
GPU TEMP67.0°C
INLET (ASHRAE)22.4°C
IT LOAD212 kW

GPU temperature forecast · R07

history P50 P10–P90 Warning ≥ 83°C Critical ≥ 90°C
OKR07 · Overheating risk score 0.08 within 1h

Proprietary R&D · Rights holder · Operator

TES Remote DCIM Platform with the GPUasS AI-Client Engine

Tech-Engineering Solutions LLP is the rights holder, R&D developer and operator of its own software ecosystem. We provide services for collecting, processing, parametric localization and predictive AI analysis of machine telemetry — and integrate client portals and smart billing on top of it.

Own source code & IP

The platform, ML models and Edge Collector are developed in-house. No third-party DCIM licences and no per-device royalties are passed on to the client.

Automated data processing services

Harvesting, normalisation into SI units, storage in time-series hypertables and predictive analysis of 25 telemetry parameters — delivered as a service from our Remote Operations Center in Astana.

Client portals & smart billing

Embedded DCIM widgets, automated report delivery and usage-based tariffing per device or telemetry volume, integrated with the client's CRM.

TECH STACK Python 3.12 FastAPI PostgreSQL 16 + TimescaleDB PyTorch scikit-learn Docker Ollama + LangChain OpenCV · Tesseract
MODULE 01 · REPORTING

Remote analytics & reporting (PDF / CSV / JSON with SHA-256)

Daily operational summaries, weekly trends and monthly branded reports with integrity hashes.

  • 10 report sections: data availability, PUE, 2D heat maps, top-10 risk devices
  • 7-day forecast, incident log and engineering recommendations
  • SHA-256 checksum on every exported file for audit trails
MODULE 02 · BITRIX24 CONNECTOR

Client portals & smart billing integration

An embeddable DCIM widget right inside the company card in the client's CRM.

  • Automatic report delivery to the client's drive
  • Tariffing per device or per telemetry volume
  • Automatic task creation on Critical alerts, escalation after 30 minutes
MODULE 03 · OCR · ANPR · QR

Remote facility security (OCR, ANPR & QR gatekeeping)

Physical access control for unattended sites without a permanent guard post.

  • Document recognition with Tesseract OCR in EN / RU / KZ
  • Automatic number-plate recognition (OpenCV ANPR) at the DC entrance
  • Cryptographically signed one-time QR passes with Anti-Passback control
MODULE 04 · LOCAL LLM · RAG

Local AI assistant with an isolated knowledge base

Local language models (Ollama + LangChain) — no data is sent to external clouds.

  • Per-tenant RAG: each data center has its own isolated index
  • Plain-language root-cause explanations for alerts (explain_alert)
  • Natural-language answers about the history of a specific facility

Why Astana · Economic efficiency

A 24/7 on-site crew for 10–50 racks does not pay off

Owners of small, modular, container and solar-powered data centers pay for 4–5 shift engineers just to watch screens at night. We replace that with automation, ML forecasting and a remote analytical center in Astana.

01

−65–75% OPEX

Radical OPEX reduction (Europe vs Kazakhstan)

Keeping a 24/7 on-site engineering crew in the EU costs €250,000+ per year in salaries and social taxes, which destroys ROI for small 10-48 rack sites or solar-powered GPU containers. Our Remote Operations Center in Astana, combined with proprietary zero-royalty DCIM software and Kazakhstan IT tax incentives, reduces monitoring costs by up to 65-75%.

02

UTC+5

Time-zone advantage of UTC+5

Astana operates in a single UTC+5 time zone, seamlessly bridging daytime operational windows from European Union (UTC+1/+2) to East Asia (UTC+8/+9) without exhausting night-shift human factor errors.

03

B2B · SLA

Tailored adaptation to your facility

We act as an intelligent analytical bridge between your physical hardware and your on-site field technicians. Under a tailored B2B contract, we configure the digital asset hierarchy (Tenant -> Site -> Room -> Rack -> Device -> Metric), calibrate ML models on your workload profiles, and deliver actionable early warnings so your technician visits the facility only when physical intervention is strictly required.

OPEX savings calculator

Compare a local 24/7 shift crew with remote AI monitoring from Astana.

12
Facility location

Estimated annual savings

€0

−0%
Local 24/7 shift crew€0
TES remote AI monitoring + on-call technician€0
Get an exact quote for my site

Indicative estimate based on average loaded labour costs (salary + employer social taxes). The final price is fixed in an individual B2B contract after a parametric audit.

On-site 24/7 crew vs TES Remote DCIM
Parameter Local 24/7 crew (EU) TES Remote DCIM · Astana
Staffing4.8–5 FTE shift engineers permanently on siteRemote analytical center + on-call field technician
Indicative annual cost, 12 racks, Western Europe≈ €324,000≈ €86,000
Detection modelReactive: alarm after a threshold is breachedPredictive: 1h / 24h / 7d forecasts with P10–P90 bands
Night shiftsFatigue and missed alarms at 3 a.m.EU night hours fall into the Astana working day (UTC+5)
Software licensingPer-device DCIM licencesProprietary platform, zero per-device royalties
Site visitsPermanent presence regardless of needOnly when physical intervention is strictly required

GPUasS AI-Client Engine v.1.0

Full control of 25 telemetry parameters and AI forecasting

Four equipment groups, all values normalised to SI units. Every metric feeds the forecasting pipeline, the anomaly detector and the composite GPU Health Index.

  • gpu.utilization%

    GPU compute core utilisation.

  • gpu.memory.usedMiBgpu.memory.utilization%

    Video memory usage and memory controller load.

  • gpu.temperature°Cgpu.memory.temperature°C

    Core and HBM memory temperature with threshold control.

    Warning ≥ 83°CCritical ≥ 90°C
  • gpu.power.drawWgpu.power.limitW

    Current power draw and configured power limit.

  • gpu.clock.smMHz

    Streaming multiprocessor clock frequency.

  • gpu.fan.speed%

    Cooling fan speed.

  • gpu.ecc.errorscount

    Corrected and uncorrected ECC memory error counter — an early indicator of chip degradation.

  • gpu.throttleflag

    Detection of thermal or power throttling events.

Mathematics & AI forecasting models

Three model families compete continuously. Automatic rolling-origin back-testing (MAE and MAPE over the last 14 days) selects the most accurate model for each metric and device group.

1 hrecomputed every 15 minutes
24 hrecomputed every hour
7 drecomputed every hour
30 dcapacity trend
01 · BASELINE

Seasonal naive & exponential smoothing

Daily and weekly seasonality. A robust reference that every advanced model must beat in back-testing.

02 · SCIKIT-LEARN

Histogram gradient boosting

Quantile loss function on lag features and rolling windows (1h, 6h, 24h) — produces the P10 / P50 / P90 confidence band.

03 · PYTORCH

Compact LSTM

Two-layer recurrent neural network: 64 units, dropout 0.1, 24-hour input window.

ISOLATION FOREST · 200 TREES / DEVICE GROUP

Anomaly detector & risk thresholds

A multivariate Isolation Forest detects unusual combinations of temperature, power and clock behaviour. When the overheating or failure risk reaches 0.60 a Warning is raised; at 0.85 an immediate Critical alert is sent.

risk < 0.60 · Normal risk ≥ 0.60 · Warning risk ≥ 0.85 · Critical
GPU HEALTH INDEX · 0–100

Composite GPU Health Index

Thermal headroom40%
No throttling25%
ECC memory errors20%
Power stability15%
AI Forecasting Pipeline INGEST TimescaleDB hypertables FEATURES lags · rolling 1h · 6h · 24h BASELINE seasonal naive · ETS HIST-GBM quantile loss LSTM 2×64 · dropout 0.1 BACKTEST rolling-origin MAE · MAPE · 14d best model ✓ ISOLATION FOREST 200 trees · anomalies FORECAST P10 · P50 · P90 | 1h 24h 7d HEALTH + ALERTS ≥0.60 WARN · ≥0.85 CRIT
AI FORECASTING PIPELINE · ingest → feature engineering → 3 model families → rolling-origin back-test → quantile forecast, anomalies, Health Index and alerts

For data center administrators & SecOps

Read-only by design. Zero inbound ports.

Technical guarantees for your system administrators and information security team. The hub physically cannot send a control command to your equipment.

01

Hub-and-Spoke & Read-Only principle

  • Edge Collector ships as a lightweight Docker container (Python 3.12, 1 vCPU, 1 GB RAM, 10 GB disk)
  • Runs as an unprivileged non-root user on a read-only filesystem (except the local spool volume)
  • No inbound listening ports and no command handlers: the hub cannot reboot a server, change GPU clocks or affect client hardware
02

Cryptographic transport & 72h local spool

  • Outbound only: HTTPS / TLS 1.2+ on port 443 to /api/v1/ingest/batch
  • Sampling every 15 s, min/avg/max aggregation over 60 s (3–6× less traffic), gzip compression
  • Every batch is signed with a per-site HMAC-SHA256 key, ±5 min replay window, strict idempotency by batch_id
  • Store-and-forward: at least 72 hours buffered locally during link outages without losing a single sample
03

Edge pseudonymization & tenant isolation

  • pseudonymize: true replaces host names (host_alias) and serial numbers with cryptographic hashes before data leaves the site
  • The mapping table stays only inside the client's DC perimeter
  • Strict per-tenant isolation of data, models and knowledge bases on the hub
Hub-and-Spoke architecture: client sites push signed telemetry to the Astana hub HTTPS :443 · HMAC HTTPS :443 · HMAC gzip · batch 60s gzip · batch 60s ASTANA HUB UTC+5 · ROC Ingestion API ML · Reports SITE-EU-01 Container GPU DC Edge Collector ● SITE-EU-02 Solar GPU Cluster Edge Collector ● SITE-EU-03 Modular Hall Edge Collector ● SITE-CIS-01 Edge Server Room Edge Collector ● ✕ HUB → SITE: NO INBOUND · NO COMMANDS
HUB-AND-SPOKE · telemetry flows one way only — from the client site to the hub; the reverse direction does not exist
Edge Collector pipeline CLIENT DC PERIMETER · READ-ONLY · NON-ROOT nvidia-smi DCGM Exporter IPMI / Redfish BMC SNMP / Modbus PDU · UPS · CRAC SAMPLE every 15 s AGGREGATE 60s min/avg/max PSEUDONYM. hash host/serial SIGN+GZIP HMAC-SHA256 LOCAL SPOOL · ≥ 72 h store-and-forward HUB INGEST API POST /api/v1/ingest/batch TLS 1.2+ · :443 outbound ✕ no reverse channel
EDGE COLLECTOR PIPELINE · sources → 15 s sampling → 60 s aggregation → pseudonymization → HMAC signing → outbound HTTPS push
# /etc/tes/collector.yaml
hub:
  url: https://hub.tesmonitoring.com/api/v1
  site_id: SITE-EU-01
  collector_id: COL-MODULAR-01
  key_file: /etc/tes/site.key        # Per-site HMAC-SHA256 secret key
schedule:
  sample_interval_s: 15              # High-frequency local polling
  transmit_interval_s: 60            # Aggregated min/avg/max push
spool:
  path: /var/lib/tes/spool
  retention_h: 72                    # 72-hour offline store-and-forward buffer
privacy:
  pseudonymize: true                 # Hash hostnames & serials at the edge
sources:
  - type: nvidia_smi
    binary: /usr/bin/nvidia-smi
  - type: dcgm_exporter
    url: http://127.0.0.1:9400/metrics
  - type: ipmi_redfish
  - type: snmp_modbus
NATIVE ADAPTERS nvidia-smi NVIDIA DCGM Exporter (Prometheus) CSV import IPMI / Redfish (BMC) SNMP / Modbus (PDU · UPS · CRAC)

How we connect your facility

Interaction protocol & contract onboarding stages

From the signed contract to 24/7 monitoring — four transparent steps.

  1. 01

    Parametric audit & hierarchy localization

    Individual contract; description of halls, racks, power inputs, cooling circuits (including solar generation and dry coolers) and notification thresholds.

  2. 02

    AI calibration on synthetic & historical profiles

    A digital load profile is generated (15 s step); gradient boosting and LSTM models are trained to target accuracy:

    MAPE ≤ 10% · power 24h MAE ≤ 3°C · GPU temp 1h Recall ≥ 0.80 · anomalies
  3. 03

    Secure Edge Collector installation (check & run)

    The container is deployed at the client site in 15 minutes with a unique site.key — without opening any inbound firewall ports.

  4. 04

    24/7 monitoring & report delivery

    Continuous metric control, instant webhook / email / Bitrix24 notifications to the on-call field engineer and monthly analytical reports.

Contacts

Request a custom protocol & SLA

Tell us about your facility — we will prepare a monitoring protocol and a contract calculation.

Please fill in this field.
Please enter a valid email address.
Please fill in this field.
Request will be sent to info@tesmonitoring.com