I design AI systems that have to survive outside notebooks: sensor streams, edge devices, CV pipelines, automation stacks, ML services, and deployment reality.
My focus is where data engineering, applied ML, embedded constraints, and systems thinking meet:
- production-ready AI and automation pipelines
- computer vision and signal processing for physical devices
- ML-backed analytics for weather, telemetry, and industrial data
- fast prototyping with Python, and performance-oriented tooling with Go
- LLM products, RAG systems, bots, and operator-facing internal tools
flowchart LR
A[Raw Sensors / Devices] --> B[Acquisition Layer]
B --> C[Validation + Cleaning]
C --> D[Feature Engineering]
D --> E[ML / CV / Heuristics]
E --> F[APIs + Bots + Dashboards]
F --> G[Operators / Clients / Devices]
E -. feedback loop .-> C
| Project | What it does | Stack |
|---|---|---|
| AI interview analysis platform | Upload-to-report workflow for interview and meeting recordings with transcription, GPT-powered analysis, bilingual reports, and persistent history | FastAPI, React/Vite, Whisper, LLM APIs, SQLite |
| Retrieval and support intelligence | Internal RAG systems over technical knowledge bases with vector search, dual-retrieval ideas, and operator-facing answer flows | Python, LangChain, Qdrant, Chroma, embeddings |
| 3D volume estimation from imagery | Photogrammetry and depth-based pipelines for point clouds, reconstruction, and object volume estimation from photos | OpenCV, Open3D, transformers, depth models |
| Industrial visual inspection | Real-time and offline CV experiments with YOLO and oriented bounding boxes for industrial object detection | Python, OpenCV, YOLO, PyTorch |
| Weather station and telemetry validation | Tools for station lookup, monthly validation, telemetry checks, and sensor-data quality control | Python, pandas, Meteostat, analytics, device data |
| Telegram automation products | Stateful bots for health, personality, and utility workflows with persistence and integration-oriented architecture | Python, aiogram, SQLite, PostgreSQL, Redis |
AI services -> model-backed APIs, bots, assistants, internal tools
Data pipelines -> cleaning, validation, feature extraction, reporting
Edge intelligence -> CV and telemetry logic near the device
Retrieval systems -> embeddings, vector stores, search, grounded generation
Voice interfaces -> transcription, chunking, speech-to-text-driven analysis
Infra glue -> Docker, Linux, automation, deployment workflows
Applied research -> turning experiments into systems that can be operated
LLM systems -> applied assistants, evaluation flows, prompt pipelines
RAG infrastructure -> embeddings, retrieval quality, support knowledge systems
Computer vision -> detection, measurement, and scene understanding for real objects
Telemetry -> weather, BLE, sensor validation, and industrial data pipelines
Automation -> bots, internal tools, Linux services, and deployment workflows
Build useful systems. Make them observable. Keep them fast. Make them survive contact with production.