Senior Data & Product Engineer
Agriculture · Geospatial · AI
I started in the field — five years as an agronomist, building yield maps from harvester sensors and writing fertilizer plans. For the eight years since, I've built the data products that run that same work: at OneSoil, as co-founder of Cropler, and now at Assist Agro, where I designed and run the company's data platform. Working products in days, running in production for years.
Most data engineers meet agriculture through a spec. I calibrated the combine and wrote the fertilizer plan — then built the systems that read the same field from space.
Agriculture is where data quietly decides whether a harvest gets in on time and whether a farm earns or loses money. I go after the problem first, and I want my work to have real-world weight — in food, land and climate.
Now — at Assist Agro and enjoying it. Open to senior and lead roles running data for large agri-food, farming and earth-observation businesses.
Range
- Satellites — multi-year Sentinel-2 productivity maps, field segmentation, management zones.
- Machinery & sensors — yield maps from harvester data, fleet capacity against field-work deadlines, crop-sensor feeds.
- Operations — plan vs. actual for every field and operation against its tech card, with probabilistic finish dates.
- Finance — crop budgets, cost per tonne, break-even yield and price, monthly cash flow, procurement plans.
- Decisions — dashboards every department works from, and AI agents that send alerts, automate accounting and edit crop plans.
Experience
Own the company's data platform end to end — built from scratch, used daily by every department from agronomy to finance and sales. Clients include some of the largest agroholdings in the region, each farming hundreds of thousands of hectares.
- 100+ dashboards that agronomy, finance and sales work from daily, fed by 80+ automated pipelines.
- Agroplanerka — a season planning and finance product for farm directors, which I designed and built.
- Schedule: weeks ahead, shows which field operations will miss their window, what each day of delay costs per crop, and which fix closes the gap — an extra combine, weekend shifts or a contractor.
- Money: before sowing, not after harvest — cost per tonne, break-even yield and price, monthly cash flow and a procurement plan, built from the farm's tech cards.
- AI agents in production — alerts, accounting automation, field monitoring, and an assistant that edits crop plans in chat.
- Geospatial pipeline — satellite remote sensing on DuckDB, field segmentation with SAM and U-Net.
- Customer-health and churn analytics that account teams use to keep agroholding clients, and dashboards that shorten the path from field data to a management decision.
- Run the platform in production — deployments, reliability, backups.
Tools: Apache Airflow, Apache Superset, DuckDB, PostGIS, Python, FastAPI, image segmentation (SAM, U-Net), AI agent frameworks, Monte-Carlo forecasting, earned-schedule analytics, 1C integration, ECharts, CI/CD, Git, Linux
Co-founded the company and took the product from zero to market. Designed it end to end with a designer and launched quickly — Superset and a Python Telegram bot delivering crop-sensor data and satellite analytics to farmers. Worked with ML engineers on phenological-stage labeling in Label Studio.
Presented Cropler at Agritechnica in Hannover — the world's largest farm-machinery fair — and built the dealer channel from the contacts made there. Owned roadmap, backlog, marketing and sales-channel expansion.
Data labeler → team lead → product manager in under a year. My agronomy background gave me an edge — I understood the users because I had been one.
Built the geospatial data pipeline for OneSoil Map: search, labeling, QA, inference support. Led the mapping team, then took over the product: user research, A/B testing, feature specs, backlog. Six months as the only PM on the platform.
Raised data completeness through targeted features. 40+ Superset dashboards used daily across all departments. Compiled the agronomic reference data that became core platform content.
Tools: QGIS, GDAL, PostGIS, Python, Jupyter, Google Earth Engine, PostgreSQL, BigQuery, Apache Superset
Three organizations across seed, feed and precision agriculture. Yield mapping with Claas, Trimble and New Holland harvester sensors; variable-rate fertilizer maps from satellite imagery; electronic field records and GIS monitoring.
Skills & Stack
Apache Airflow, Apache Superset, DuckDB, PostgreSQL / PostGIS, Python, FastAPI, ETL pipelines, data modeling
BigQuery, ECharts, ad-hoc analysis
AI agent frameworks, LLM integration, agent memory & tools, MCP servers
Speech AI (Whisper, speaker diarization), Telegram bots
Sentinel-2 time series, management zones & VRA maps, GDAL, QGIS, image segmentation (SAM, U-Net)
rasterio, GeoPandas, STAC, Google Earth Engine, yield mapping & combine calibration, MapLibre / PMTiles
Product strategy, product analytics, B2B & B2C, crop economics & budgeting
A/B testing, user research, backlog prioritization
CI/CD, Git, Linux, Docker, server administration, deployments, backups
Russian (native), English
Precision agriculture, variable-rate application, on-farm trials, crop monitoring, phenology, agrochemistry, farm operations & finance
Side Projects
I keep a few products running on my own time — to stay hands-on across the whole stack and to learn how cheaply and reliably a product can run. All are live; none is a business.
Field polygon in, prescription out: a multi-year Sentinel-2 productivity map, management zones with inter-year stability, control strips for on-farm trials, and a split fitted to the farm's machinery — or none, if the field is uniform. Harvester yield data then scores the trial strips against the zones. 2,500+ production runs — private for now.
Turns raw precision-agriculture data into interactive maps: a self-written yield-processing engine (cleaning, calibration, zone delineation) behind an API, streaming of field archives up to 4 GB, and PMTiles served to a MapLibre client. 8 languages — mapagtech.pilip.io
Built for my own MTB marathon riding, now a community route map: a multi-profile routing engine (road / gravel / MTB / walking), a route builder with off-road drawing and ring-route generation, elevation profiles, a ridership heatmap and device-tuned export for Garmin, Wahoo, COROS and Karoo — nakat.bike
Transcription with speaker diarization — who said what, recognized across recordings, 90+ languages — on my own self-hosted Whisper inference API. The Telegram bot holds 1K+ MAU (2K at peak) with zero ad spend — voxprint.app · @VidVKYT2AudioBot
Also: Inkover (AI manga translation), Rent-Stat (rental-market analytics), EvAuto (EV catalog maintained by an agent pipeline).
Education
Workforce Management in the Agricultural Sector (second degree)
Technology for Storage and Processing of Plant Raw Materials
Off the Clock
MTB marathons on forest trails — the reason Nakat exists — and épée fencing.
What's Next
Senior or lead work where data is the product or the backbone — agriculture, food, earth observation, climate. I've co-founded a startup and built products alone; next I want the other end of the scale: large systems, large organizations, strong teams.
Open to conversations — and happy to talk shop about satellites, yield data or farm economics, hiring or not.