Senior Research Engineer
GeoAlert
Geoalert develops Mapflow, a geospatial AI platform that uses satellite and aerial imagery to map features such as buildings and vegetation.
Feb 2026 — PresentBuilds and optimizes SAM 3 and remote sensing models for Mapflow, with reported gains in segmentation F1, inference speed, and model size.
• Extended SAM 3 with a polygon prompt by replacing box ROI-align with a masked variant in the prompt encoder, improving object F1 on non-rectangular, border-aligned and closely packed objects with no retraining.
• Built the SAM 3 interactive inference service for Mapflow: split the model into four components carried through ONNX and TensorRT to Triton on V100 hardware in fp16 with fp32 wrappers on numerically sensitive layers, plus session storage, image-embedding caching and fast decoding; approximately 50% speedup and VRAM reduction.
Machine Learning Engineer
Faceplate Industrial AI
Feb 2024 — Feb 2026Led AI delivery for an energy client, building SCADA alert summarization and forecasting systems, with reported forecast MAPE below 10% and 5%.
• Built an LLM, RAG and human-in-the-loop pipeline converting high-volume SCADA alerts into incident summaries and action plans: time and space grouping by windowing and graph adjacency, LLM summarisation, retrieval from dispatcher guidelines and live telemetry.
• Designed automated offline evaluation of summary and action quality using a larger LLM as judge; tracked latency and the before/after ratio of messages reaching dispatchers.
Research Engineer
Huawei
Huawei is a global telecommunications and technology company that develops network infrastructure, consumer devices, and enterprise products.
Jun 2019 — Feb 2023Worked on five Wi-Fi research projects, two reaching product deployment, and developed algorithms, testbeds, and software across simulation and firmware.
• Wi-Fi R&D across five research projects, two of which were taken through to deployment in shipping products; progressed from intern to researcher.
• Developed an adaptive A-MPDU packet-aggregation algorithm for access points, deriving the throughput model that trades header overhead against receiver equaliser ageing, and validating it in C++ and Python simulation.
• Co-developed predictive MIMO pre-coding on the Grassmann manifold using a two-point geodesic auto-regressive model, implemented in Matlab and evaluated on real channel data.
Engineer
Xumanless
Jan 2019 — Jan 2020Helped ship an automated 1,000-plus-place coat room, building its software, electronics, and PCBs and installing the racks.
• Joined the company at an early stage and helped ship a 1000+ place automated self-service coat-room, still in operation at a Moscow university.
• Built the upper-level business logic in Python (ID card to locker matching) and the lower-level logic in C (magnetic lock control, communication protocols); designed the two-sided PCBs in Altium, soldered and debugged the boards, and installed the racks.
Laboratory Assistant
NSU / Institute of Automation and Electrometry SB RAS
A joint Novosibirsk State University and Institute of Automation and Electrometry (Siberian Branch, Russian Academy of Sciences) group.
Jan 2017 — Sep 2018Built hardware and software for fibre optics research systems and a fibre-optic bending sensor, and contributed to two publications.
• Fibre-optics research in the Laboratory of Nonlinear Optics of Guided Wave Systems, a joint Novosibirsk State University and Institute of Automation and Electrometry (Siberian Branch, Russian Academy of Sciences) group.
• Built parts of a fibre Bragg grating inscription system in hardware and C++/Python: microcontroller-controlled tilted-grating inscription, a laser beam tilt control system, and a test-bed for the produced gratings.
• Developed the hardware, firmware and UI for a fibre-optic bending sensor; contributed to two peer-reviewed publications in fibre optics.
