What we've shipped.

The operations team was drowning in alerts. Across hundreds of sites, dashboards generated thousands of notifications per day, but only a small fraction were truly actionable. Operators spent 60%+ of their time triaging noise instead of fixing problems. Standard BI tools couldn't help because the real questions were ambiguous: 'Why is this string under-performing today? Weather, soiling, a faulty inverter, or wiring?' Off-the-shelf chat-with-your-data tools hallucinated over operational data — unacceptable in a regulated industrial setting.

An agent-native reasoning system that grounds every recommendation in real time-series data, asset metadata, and physics-based calculations. The LLM accesses these via typed tool calls — not vector search over text. Every answer cites its sources, every interaction is audited, and sensitive data never leaves the customer's control via a single-inference-and-destroy pattern.

Existing gateway firmware was fragmented — one binary per deployment, no shared platform, every protocol bolt-on was a copy-paste regression risk. Each new inverter brand was a two-week regression cycle. OTA stories devolved into 'we ship engineers to site every quarter.' The customer needed a productized firmware OS, not another one-off custom delivery.

A productized Linux-based Edge Firmware OS for industrial gateways. Runs on ARM Cortex-A53 reference hardware and on customer-owned third-party SKUs (multi-tenant runtime licensing). Single firmware tree handles 8+ protocols, identifies devices automatically on power-up, and updates via signed A/B OTA with staged rollout.

The operator managed assets from 8+ inverter brands (Huawei, Sungrow, Growatt, GoodWe, et al.), each with proprietary APIs and undocumented Modbus dialects. They needed a single platform that ingests, normalizes, correlates string-level + module-level + weather data in real time, and surfaces under-performing assets fast enough for the O&M team to act before revenue loss compounded.

End-to-end IoT platform: edge data collectors deployed on-site, time-series ingestion at multi-GWp scale, multi-tenant cloud platform with row-level access for asset owners / O&M / regulators, and an operations dashboard with auto-generated work orders that both humans and cleaning robots can action.

The bank operates a global regulatory file-flow and KPI calculation platform feeding both internal risk teams and regulator-facing reports. Development had to coordinate across four geographies: business owners in New York, architecture in Northern Ireland, development in Shanghai, QA in India. Every requirement clarification, code review, and deployment was async + scheduled English call. Compliance windows meant zero production downtime tolerance.

Acted as Solutions / Delivery Lead in the Shanghai team, owning requirement analysis between NY business and our delivery org. Wrote production code (Java + WebLogic + Flex), coordinated 7×24 production handoff between Shanghai (day) and NY (night) shifts, and shepherded 50+ JIRA tickets + 30+ design docs through compliance review.