Systems thinking
I start with data contracts, lineage, failure modes, and operational cost instead of treating pipelines as scripts.
Tokyo, Japan / Geospatial Big Data Engineer
I build geospatial data platforms that turn mobility signals into reliable decisions.
Big data engineer focused on GPS-scale pipelines, spatial analytics, and production data products for smart cities, mobility, and urban planning teams.
I design systems that ingest high-volume location data, validate it, enrich it with spatial context, and ship it as analytics-ready datasets, dashboards, maps, and APIs.
PythonSparkKafkaAWS S3AWS GlueClickHouseKubernetesH3GeoPandasETL/ELTREST APIsOpenADR
8+ yrs
data / ML systems
GPS-scale
mobility analytics
AWS + Spark
cloud data platforms
Tokyo
Japan-based
About
I solve the messy middle between raw mobility data and decision-ready geospatial products.
My work sits at the intersection of distributed data engineering and geospatial analytics. I care about the parts that make data products dependable: ingestion contracts, partitioning strategy, spatial indexing, reproducible transformations, cost-aware cloud execution, and clear outputs that technical and non-technical teams can trust.
Spatial fluency
I work with coordinates, grids, joins, time windows, OD movement, road volume, and POI footfall as first-class data concerns.
Product mindset
I care about the final consumer: analysts, researchers, consultants, dashboards, APIs, and map-based workflows.
Case studies
Systems work framed by problem, solution, stack, and impact.
These are public-safe summaries of real production patterns: ingestion, spatial enrichment, serving layers, optimization, and operational workflows.
LocationMind / GPS-scale geospatial pipelines
Mobility Data Foundation
Raw GPS
->Validation
->Spatial enrichment
->Feature tables
->Analytics outputs
Problem
Raw mobility inputs are noisy, high-volume, and difficult to reuse across research, consulting, analytics, and product workflows.
Solution
Designed reproducible ingestion and transformation layers that validate, normalize, aggregate, enrich, and prepare location records for downstream analytics.
Impact
- Reduced repeated preprocessing work by standardizing analytical datasets.
- Improved confidence in outputs through validation, documentation, and repeatable transformations.
- Created foundations for maps, dashboards, CSV exports, and model features.
PythonSparkAWS S3AWS GlueGeoPandasH3Data quality
LocationMind / Urban planning and location intelligence
People-Flow Analytics Products
GPS events
->Spatial indexing
->OD / volume / footfall
->Maps
->Stakeholder delivery
Problem
Urban and mobility teams need reliable views of movement patterns without manually interpreting raw GPS trajectories.
Solution
Built analytical outputs for OD movement, road-volume intensity, and POI-footfall patterns using spatial aggregation and map-ready data products.
Impact
- Translated movement signals into decision-ready geospatial outputs.
- Supported stakeholder workflows across research, consulting, and product teams.
- Made repeated mobility analysis easier to reproduce and compare over time.
PythonH3GeoPandasClickHouseMapsCSV products
GridSolutions / Optimization and platform integration
Energy Pricing Optimization System
Business inputs
->Optimization module
->API layer
->Client systems
->Operations
Problem
Pricing workflows needed optimization logic that could be integrated into operational systems and maintained over time.
Solution
Implemented optimization modules, API interfaces, and OpenADR-related integrations while improving architecture reliability.
Impact
- Moved optimization research closer to production workflows.
- Created integration surfaces for client platforms.
- Reduced operational risk by refactoring away single points of failure.
PythonMINLPREST APIsOpenADROptimizationSystem refactoring
Smart Data Solutions / OCR and operational automation
Document Intelligence Pipeline
Scanned claims
->OCR
->Classification
->Entity extraction
->Review interface
Problem
Scanned claims required structured extraction and classification before they could move efficiently through operational workflows.
Solution
Combined OCR extraction, text classification, rule-based extraction, NER, and interface improvements to support faster manual review.
Impact
- Improved the path from scanned documents to structured operational data.
- Reduced friction for manual keying workflows.
- Connected ML extraction with practical back-office usability.
OCRTesseractFineReaderWEKARandom ForestNER
System design
Architecture patterns for geospatial big-data systems.
The site now exposes architecture thinking directly instead of hiding it inside tool lists.
Mobility Data Lakehouse
A reusable pattern for turning high-volume GPS data into analytics-ready spatial features and product outputs.
Sources
->Kafka / batch ingest
->S3 raw zone
->Spark + Glue
->H3 feature marts
->ClickHouse / APIs
Schema contractsPartition-aware processingSpatial indexingCost-aware computeObservable outputs
Geospatial Product Layer
A serving layer designed for teams that need fast comparison across movement metrics, time windows, and geographic cells.
Feature tables
->Aggregation jobs
->ClickHouse
->Map tiles / CSV
->Dashboards
->Decision workflows
Low-latency readsReproducible exportsMap-ready geometryClear lineageStakeholder trust
Skills
Grouped by the work they enable, not by tool names.
The stack matters because it supports large-scale movement data, spatial context, reliable pipelines, and fast serving layers.
Data Engineering
Designing batch and streaming pipelines with explicit data contracts, validation, partitioning, orchestration, and observability.
PythonSparkKafkaETL/ELTData qualityTime-series features
Cloud & Infra
Shipping data products on cloud infrastructure with cost-aware execution, containerized services, and production deployment habits.
AWS S3AWS GlueEC2KubernetesClickHouseMonitoring
Geospatial
Turning raw mobility traces into spatially indexed, map-ready analytics for people-flow, OD movement, road volume, and POI footfall.
H3GeoPandasSpatial joinsMobility dataMapsLocation intelligence
Experience timeline
A career arc from ML systems to geospatial big-data platforms.
The through-line is production work: data quality, integration, cloud systems, operational reliability, and useful outputs.
Building mobility and geospatial data pipelines that transform raw GPS-scale inputs into people-flow analytics, map-ready outputs, and decision products.
- Design ingestion, validation, normalization, aggregation, and geospatial enrichment workflows for mobility datasets.
- Build time-series and spatial features for people-flow analysis, dashboards, maps, and CSV data products.
- Improve reproducibility, monitoring, performance, documentation, and governance across the data lifecycle.
PythonSparkAWSGeospatial joinsTime-series featuresData quality
Delivered optimization and integration systems for energy pricing workflows, including business logic, APIs, and OpenADR VEN projects.
- Implemented optimization modules using Mixed-Integer Nonlinear Programming for pricing strategy workflows.
- Built API interfaces for client-platform integration and maintained OpenADR VEN projects.
- Refactored architecture to reduce single points of failure and improve operational reliability.
PythonOptimizationMINLPOpenADRAPI designReliability
Led delivery of machine-learning-backed APIs and cloud services, bridging implementation architecture with client-facing product needs.
- Led end-to-end project lifecycles from planning through production delivery.
- Collaborated with architects to define implementation plans aligned with client constraints.
- Developed REST APIs and cloud integrations that exposed ML functionality to applications.
Machine learningREST APIsAWS LambdaAmazon EC2Team leadership
Jul 2018 - Nov 2019
Machine Learning and OCR Engineer
Smart Data Solutions / Full-time
Kathmandu, Nepal
Built OCR and document-intelligence workflows for scanned claims, combining extraction, classification, and operational interface improvements.
- Extracted characters from scanned claims using Tesseract, FineReader, and Cartouche.
- Classified extracted text with Random Forest models and extracted entities using rules and NER.
- Improved manual keying interfaces to reduce operational friction.
OCRTesseractFineReaderWEKARandom ForestNER
Tech blog / insights
Technical themes behind the work.
Areas I keep sharpening through production work, architecture decisions, and practical geospatial data-system design.
Spatial indexing
Designing H3 Feature Tables for Mobility Analytics
How to choose resolution, partitioning, and aggregation boundaries when GPS events need to become reusable spatial features.
Data reliability
Data Quality Checks That Matter for GPS Pipelines
Practical validation patterns for timestamp consistency, coordinate sanity, duplicate trajectories, and downstream trust.
Serving layer
ClickHouse for Geospatial Data Products
Where columnar serving fits after Spark transformations, and how to think about query shape, rollups, and product latency.
Education
Computer science foundation for systems, data, and applied ML.
Formal computer science training supports the later career arc across software systems, data platforms, machine learning, and cloud applications.
Nepal
BSc CSIT
Deerwalk Institute of Technology, Tribhuvan University
Bachelor of Science in Computer Science and Information Technology from Deerwalk Institute of Technology, affiliated with Tribhuvan University, grounding later work in software systems, data platforms, and applied machine learning.
Contact
Open to geospatial data platforms, mobility analytics, and systems-heavy engineering work.
Reach out if the problem involves high-volume data, spatial analytics, distributed pipelines, or turning complex data into something a product team can operate.