Weather, festivals, industrial cycles — Indian demand is exceptionally volatile. Generic models miss it badly.
Millions of meters streaming half-hourly data. Storage is cheap; useful inference is not.
Linemen, water-board surveyors, gas inspectors — they need apps that work on rugged Android, in vernacular, offline.
Transformers, feeders, pipes — most utilities track them in Excel. Real intelligence comes from sensor + maintenance + outage data fused.
Hybrid time-series + LLM-prompted feature pipelines. Calendar, weather, industrial-load features tuned for India. MAPE under 4%.
Smart-meter ingestion at scale, anomaly detection on consumption + voltage, theft-pattern flagging. Wired into existing SCADA / OMS.
Lineman / surveyor / inspector apps. Offline-first, multilingual, voice-driven. Photo-evidence + GPS + asset-tag scanning built in.
Sensor + CMMS + outage fusion. Failure-mode tagging. Predictive maintenance on transformers, feeders, distribution lines.
ML-based forecasting on a 14-state discom dataset. 3.7% MAPE on day-ahead, replacing a 9% statistical baseline.
Offline-first Flutter app for linemen across 4 states. Voice-driven UI, photo-evidence + GPS asset tagging. 8K linemen onboarded in 8 weeks.
Yes. Read-only integration via OPC UA / IEC 61850 / vendor-specific adapters. We never write back to OT systems unless an explicit, audited bridge is defined.
Yes — most discom and water-board engagements are on-prem or sectoral-cloud (NIC, state DCs). We deploy via Terraform / Ansible with air-gap support.
We've handled 4M+ meters at half-hourly resolution. Storage on cost-tiered Postgres + object storage; analytics on a thin OLAP layer.