Research, R&D and digitalisation
From research directions to applied AI solutions — medicine, EdTech, hydropower digital twins, GovTech. Open for student participation and international collaboration.
What the department works on
automation of technological processes and smart manufacturing
robotic and mechatronic systems
applying AI methods in control and diagnostics
digitalization of public administration processes (GovTech)
MedAI KG
AI analysis of X-rays and CT scans: the model localises pathologies and generates a preliminary report in seconds.
Acute shortage of radiologists in rural Kyrgyzstan and Central Asia. Patients wait 2–5 days for a scan reading. Early diagnosis of TB, pneumonia and fractures is critical.
Web service: doctor uploads DICOM or JPEG → ResNet-50 returns a heat-map with pathology probabilities + brief report. Final diagnosis is made by the doctor.
- ML engineer (ResNet training)
- Python / FastAPI backend
- React frontend
- Medical consultant
HealthBot КГТУ
Telegram triage bot: describe symptoms — get a differential list and urgency recommendation.
Most residents of Kyrgyzstan do not know which specialist to see at first symptoms. Emergency rooms overloaded with non-urgent cases that could go to a clinic.
Claude API + RAG on ICD-10 and Ministry of Health KR clinical protocols. The bot asks clarifying questions, builds a list of likely diagnoses and assigns urgency (green / yellow / red).
- LLM / Prompt Engineer
- Python Backend
- Medical editor (Ministry of Health protocols)
ClinicalNLP KG
NLP annotation of medical records in Russian and Kyrgyz: free-form doctor text → structured diagnostic entities.
Medical records in Kyrgyzstan are kept as free text. Impossible to automatically analyse data for epidemiology, physician training or AI research.
Fine-tune spaCy / LLaMA on anonymised records corpus. Model extracts symptoms, diagnoses, medications, dosages, dates. Data exported to HL7 FHIR standard.
- NLP engineer
- Annotators with medical background
- Partner clinic for corpus
КГТУ Brain
Department RAG assistant: answers questions about programmes, schedule and learning materials in dialogue mode, available 24/7.
Students and applicants spend hours searching for department information. Faculty overloaded with repetitive questions about the study process.
LightRAG graph of 600+ department concepts + voice interface (TTS) + website integration. Escalates complex cases to a live operator.
- Content editor (knowledge base)
- React frontend
- DevOps (deployment on aios.kg)
CodeTutor KG
AI programming mentor in Russian and Kyrgyz: explains errors, teaches algorithms, generates practice tasks.
1st–2nd year students lack individual attention when learning Python and C++. No quality teaching materials in the Kyrgyz language.
Web IDE with a built-in LLM mentor. Student writes code → mentor explains errors in Russian/Kyrgyz, suggests the next step, generates a similar task for reinforcement.
- React / Monaco Editor frontend
- Prompt Engineer
- Curriculum designer (year 1–2)
- Kyrgyz translator
HydroTwin KG
Digital twin of a small HPP: flow simulation, output forecast, intelligent equipment monitoring.
Over 50 small HPPs in Kyrgyzstan operate without digital monitoring. Failures and under-generation from late fault detection. No flow forecasting tools.
Digital twin: flow model (hydrological data + ML) + SCADA telemetry + 72-hour generation forecast dashboard. Pilot at a Naryn Oblast small HPP. Co-developed with ZJWEU.
- Hydrology / ML engineer
- Embedded / SCADA developer
- HPP operator partner
Glacier AI
Glacier melt forecast for Kyrgyzstan via Sentinel-2 satellite imagery: ML model on a unique regional dataset.
Kyrgyzstan has lost over 20% of glacier area in 50 years. Agriculture, small HPPs and water supply depend on glacial river flow. No accurate forecasting tools for Central Asia.
Pipeline: Sentinel-2 download → glacier segmentation (U-Net) → area dynamics time series → seasonal flow forecast. Public dashboard with maps.
- ML / Computer Vision (U-Net, satellite data)
- GIS specialist
- Climatologist / hydrologist consultant
AgroIoT KG
Smart irrigation for Kyrgyzstan: IoT soil-moisture sensors + ML water-need prediction + automated irrigation control.
65% of Kyrgyzstan water used in irrigation, with 35–40% lost to over-watering. Small farmers lack access to precision agriculture tools.
Testbed: 4–6 ESP32 sensors (moisture, temperature, precipitation) → MQTT → InfluxDB + TFLite ML forecast + web dashboard + SMS alerts to farmer. Pilot at KSTU teaching farm.
- Embedded / ESP32 developer
- ML on microcontrollers (TFLite)
- IoT backend (MQTT, InfluxDB)
- Agronomy consultant
Automating KR ministries and agencies
Student project teams
Under faculty guidance they run pilot automation projects for ministries and agencies: process analysis, electronic document workflow, "Tunduk" integration, dashboards and reporting.
AI tools in research
Preparing analytical and design documentation (specs, process surveys, reports) with AI under faculty supervision.
Contract work
Automation work for government bodies and enterprises on a contractual basis under the KSTU charter and KR legislation.
- annual student scientific and technical conference
- participation in national and international competitions and exhibitions
- publication of the best student works
- preparation for robotics competitions
List of faculty publications from the last 3 years and active grants — in preparation.
Section updated annually. An empty or stale section is worse than none — we publish only verified data.
Joint Lab with China
Zhejiang University of Water Resources and Hydropower Engineering (Hangzhou) — partner in AIoT, HPP digitalisation and machine learning. Joint publications, BRI/SCO grants, student exchange.