Sam Field

Solo consultancy. Contract integration engineering for process manufacturing and utilities clients.

• Built a .NET ingestion service reading from a PI historian for a specialty chemicals client, covering roughly 180,000 tags and replacing a nightly CSV export with streaming reads — analysis latency went from 19 hours to under a minute.
• Built a reconciliation harness that replays a historian window through both the legacy export and the new pipeline and diffs sample-for-sample, including quality codes — the tool that actually got the migration signed off.
• Diagnosed a long-standing “phantom gap” defect at a water utility — an upstream device emitting timestamps in local time across a DST boundary, silently discarded by the ingestion layer. Wrote the detection check now run against every new data source.
• Ported a Python ingestion prototype to a supervised .NET agent with bounded-memory buffering and at-least-once replay, sustaining ingest through a 6-hour network partition with zero sample loss.
• Acted as the escalation engineer for three clients’ integration stacks, including on-call for plant-hours incidents.

Mid-size vendor of asset-performance software for marine and port operations.

• Owned the shared integration framework behind every first-party data source — lifecycle, credential handling, retry/backoff, schema mapping and observability — adopted across nine source systems.
• Reduced steady-state ingestion cost by 62% by replacing per-tag polling with OPC UA subscriptions and deadband filtering, without loss of fidelity on step-change signals.
• Introduced a protocol simulator that reproduces historian misbehaviour (duplicate timestamps, quality flapping, mid-query disconnects) in CI; integration escapes fell by roughly two thirds over the following year.
• Led the security review and remediation of the on-premise agent — moved from stored plaintext credentials to OS keystore integration and short-lived tokens, and removed an outbound-port requirement that had been blocking three enterprise deals.
• Mentored four engineers; ran the internal “time is not a number” workshop on time-series pitfalls, now part of onboarding.

Systems integrator serving pulp, paper and food processing clients.

• Built and deployed OPC UA and Modbus collectors across eleven plant sites, several with no reliable WAN, using store-and-forward buffering sized to survive a 72-hour outage.
• Implemented asset-model mapping that translated inconsistent per-site tag naming into a common hierarchy, making cross-site comparison possible for the first time.
• Worked directly with process engineers on-site to validate that ingested series matched control-room behaviour — the habit that shaped how I test everything since.
• Built SQL ingestion pipelines landing plant data into a central SQL Server warehouse on a fixed schedule, with idempotent loads so that a re-run after a failed night could not double-count.
• Wrote the commissioning runbook used for subsequent site rollouts.

Data-acquisition hardware and companion software for environmental monitoring.

• Developed firmware-adjacent acquisition software in C and C# for multi-channel sensor units, including calibration and drift compensation.
• Built the desktop tooling field technicians used to configure and diagnose deployed units, cutting typical site-visit time roughly in half.