James Westover

Westover Labs / Neher Data Services — Berlin, Germany

Founder & Principal Architect · 2024 – Present

• Architect and operate a cloud-native microservices platform emphasizing zero-trust networking: Cloudflare Tunnels for public API ingress, Tailscale mesh VPN for service-to-service and CI/CD access, and Ansible for fully declarative infrastructure management — eliminating manual configuration drift.
• Design and maintain a PostgreSQL knowledge graph with pgvector (HNSW index) for high-dimensional semantic search, serving both REST and MCP API surfaces. Implemented L0/L1/L2 dictionary compression via PostgreSQL triggers and a continuous compaction worker with semantic deduplication (cosine similarity >0.92 threshold).
• Build FastAPI inference services with ONNX Runtime — 10x latency improvement over HuggingFace-based inference. Designed tiered authentication (free/paid), per-tier rate limiting, and token exchange system backed by Apple IAP validation, using SQLite for token persistence.
• Pioneer AI-augmented systems development — orchestrating 12 specialized AI agent personas with intent routing, Jira workflow automation, and persistent knowledge graph context, demonstrating team-level architecture and delivery throughput as a solo principal.
• Design Bayesian sports modelling system using PyMC5, Polars, and DuckDB — probabilistic optimization pipeline with full data engineering from ingestion through model output.

URBN — Philadelphia, PA (Remote)

Software Engineer · September 2024 – January 2025

• Maintained Django application managing URBN's internal product catalog across multiple retail brands (Urban Outfitters, Anthropologie, Free People).
• Contributed to asynchronous task processing and event-driven workflows using Celery and Kafka.
• Improved build systems and developer tooling for enhanced engineering velocity across the platform team.

Uber — San Francisco, CA / Remote

Staff Software Engineer — Navigation · February 2021 – October 2022

• Senior technical contributor on Uber's navigation platform — one of the largest distributed systems in the company at ~10,000 nodes per data center, serving all routing and mapping globally. Code continues to run on 30M+ driver devices.
• Architected support for off-route cross-street links in the route object — reduced p95 off-route detection latency from ~5 minutes to ~2 minutes through improved heuristics and on-device data refresh.
• Led critical incident response for GPS clock offset bug affecting data integrity across millions of devices globally — diagnosed root cause, coordinated cross-team remediation at scale.
• Redesigned Python-based simulation infrastructure using multiprocessing parallelization, scaling test capacity from 500 to 10,000 trips per run — enabling high-fidelity regression testing for route deviation algorithms.
• Led Postmates acquisition technology integration: compliant data deletion strategies across distributed databases (Go microservices, Schemaless datastore, identity systems).

Senior Software Engineer — Python Platform (Office of CTO) · January 2019 – June 2020

• Founding member of the platform team driving Uber's enterprise-wide Python 2→3 migration — spanning hundreds of teams and 90%+ of production services.
• Built dependency graph tooling mapping the entire Python microservice ecosystem; automated migration targeting top 200 repositories and achieved 93% service conversion.
• Major contributor to tchannel-python — core Tornado-based Thrift RPC transport library — including full Python 2→3 migration of the transport layer shared across hundreds of services.
• De facto Python platform support for Uber's data science organization: notebook infrastructure, distributed query systems, Airflow fork maintenance.

Software Engineering Manager — Location Platform · February 2018 – January 2019

• Founded and architected Uber's Location Platform from scratch — designed GDPR-compliant global location data ingestion pipeline for the entire driver and rider fleet.
• Drove global GPS+sensor data ingestion reliability from ~80% to 97%+, directly supporting RideCheck safety initiative (crash detection and emergency response).
• Owned architecture strategy for globally distributed Cassandra at 2–10M QPS: consistency model selection (QUORUM, LOCAL_QUORUM, ONE) per read/write path based on latency vs. availability trade-offs, multi-region deployment, tombstone management, and capacity planning.
• Launched ShadowMaps geospatial correction project — multipath-aware GPS correction using building geometry to improve urban location accuracy.
• Directed anti-fraud architecture for GPS location spoofing detection — program delivered $30M+ in annual savings.
• Built Location Platform team from scratch: key hires, org structure, technical roadmap, and cross-team alignment with data, platform, and safety organizations.

Postmates Inc. — San Francisco, CA

Senior Software Engineer 2 — Core Infrastructure · June 2020 – February 2021

• Led monorepo-wide Python 2→3 migration across 1M+ lines of code — navigating 3 coexisting Django versions, deep bytes-vs-unicode incompatibilities, and performance-critical delivery marketplace logic.
• Delivered 4x CI build time improvement by orchestrating distributed test execution across Kubernetes pods — 40–50 minute feedback loops reduced to 10–15 minutes, directly unblocking 100 engineers.
• War room participant during peak-traffic scaling incidents — real-time incident response alongside infrastructure and SRE leadership.
• Partnered with infrastructure on Envoy service mesh adoption and OpenTelemetry distributed tracing integration across the microservice fleet.
• Lived through the full AWS → GCP cloud migration. All Kubernetes work was CLI and YAML-first — kubectl, k9s, deployment manifests. Never touched the console. StatefulSets for databases inside k8s were the genuinely hard part.

HERE Technologies — Berlin, Germany

Software Engineering Manager → Senior Software Architect · April 2015 – February 2018

• Architected lambda data architecture for vehicle fleet telemetry: Apache Storm (Python bolts) as the real-time streaming layer, Kafka as the ingestion backbone, Elasticsearch for search and analytics, and PostgreSQL/PostGIS for persistent geospatial storage — processing 250M+ GPS events/day from global mapping fleets.
• Ran what was likely the largest single unsharded PostgreSQL/PostGIS instance in production — the entire HERE world map in one database, pre-HDFS-GIS era. The architectural hard problem: spatial queries crossing APAC/EMEA/LATAM/NORAM data partition boundaries for autonomous vehicles driving in real time. Invented the approach because the tooling hadn’t caught up yet.
• Used Apache Thrift as universal IDL across the full data stack: one schema generated Python services, C++ ROS sensor nodes, and JS frontend types. Storm topology with Thrift-classified tuples gave the streaming layer a typed contract end to end.
• Used GeoServer for tile generation pipeline: raster source data → vector tiles → custom Thrift objects consumed downstream by the fleet visualization stack.
• Built “Weatherman” event-driven orchestration system for dynamic map reprocessing workflows — Kafka-coordinated batch processing enabling quality assurance pipelines across downstream consumers (Fleetvision 2, Spotcam Monitor, HT-Route Creator).
• Designed Tornado-based microservice APIs for real-time fleet management, integrating GPS, IMU, and camera telemetry streams with React-based frontends including a 16-display video wall in Berlin HQ.
• Led ADAS prototyping pipelines integrating Python processing with ROS for rapid prototyping of sensor fusion and localization — including sensor-to-cloud architectures built with Intel, Nvidia, Audi, and Mercedes-Benz.
• Designed software abstraction for FPGA-based JPEG-compressed sensor streams (HT4 next-generation mapping rig) — hardware-software integration for reliable cloud ingestion of camera and sensor data at scale.
• Led ISO 9001 certification for safety-critical software governing mapping vehicles on public roads — authored QMS documentation, led auditor engagement, implemented process quality standards.
• Scaled engineering organization from 2 to 15 engineers across 5 subteams (backend, frontend, mobile/tablet, DevOps, embedded); maintained Jenkins + Ansible CI/CD pipeline.
• Built HT-Route Creator: replaced legacy Windows-only fleet routing tool with a cloud-hosted web application with pluggable routing engine APIs — improving consistency of generated routes across all fleet managers globally.

Sonobi — New York, NY

Software Engineer · December 2013 – January 2015

• Designed and built RabbitMQ-based distributed job processing system in Python (multiprocessing + multithreading) for analytics across 250M+ daily ad impressions.
• Architected ETL data warehousing pipeline using S3, Redshift, and Python for campaign analytics with robust logging and validation.
• Designed probabilistic group overlap estimation system using HyperLogLog and MinHash sketches with Redis for storage and NumPy for numerical computation — enabling efficient large-scale audience intersection without storing raw user data.
• Explored streaming analytics with Kafka, Logstash, and Hadoop for log-level event processing.

University of Rochester — Rochester, NY

Assistant Systems Administrator · October 2006 – October 2012

• Built and administered Beowulf-style HPC clusters using ROCKS, MPICH2, and OpenMPI for Monte Carlo simulations and molecular dynamics (LAMMPS, VASP).
• Constructed private cloud infrastructure on Eucalyptus (open-source EC2 equivalent), later migrated to VMware — early practical experience with virtualization and cloud compute abstractions.
• Designed real-time microsecond-precision timing systems for laser cooling in Bose-Einstein condensate experiments — hardware-software integration with strict latency requirements.