Our client, Zenith Renewables, operates some of the largest distributed solar farms across multiple states, managing over 500 MW of capacity. As their asset base grew, so did their operational complexity and costs. They struggled with:

• Minimal predictive capabilities, relying on spreadsheets and manual forecasts for energy yields, leading to frequent mismatches between expected and actual production.
• Maintenance teams working reactively, often dispatched only after noticeable generation drops, increasing downtime and reducing overall returns.
• No standardized way to track carbon offset data or ESG metrics, limiting their ability to attract sustainability-focused investors.

Zenith envisioned an AI-first platform that would:

• Predict hourly and daily energy production with high accuracy.
• Automate fault detection and proactively schedule maintenance before problems escalated.
• Generate comprehensive ESG reports to support green financing and regulatory compliance.

They chose ThoughtWin IT Solutions for our unique ability to combine high-throughput IoT data engineering, machine learning, and secure, multi-tenant cloud platforms.

• High-throughput Ingestion Pipeline: Built a pipeline with MQTT brokers feeding into TimescaleDB, handling millions of data points daily with sub-second latency.
• AI Prediction Models: Developed AI models in Python (TensorFlow, Scikit-Learn) to predict hourly generation within ±5%, learning from weather APIs, past performance, and degradation patterns.
• Anomaly Detection Microservices: FastAPI microservices continuously scanned for anomalies, auto-generating work orders to the client’s existing ticket system.
• Secure Authentication: JWT + OAuth2 secured both device feeds and multi-tenant user dashboards.
• Scalable Deployment: Deployed on AWS EKS with self-healing, autoscaling pods and encrypted at-rest storage.

• TimescaleDB over Cassandra: Comparative load tests showed Timescale delivered superior JOIN and aggregation speeds essential for multi-site, multi-inverter dashboards and ESG rollups.
• Dedicated ML Inference Nodes: Opted for isolated GPU-enabled EC2s instead of embedding models inside the ingestion pipeline, ensuring data flows never stalled due to model compute.
• Event-driven Maintenance Triggers: Designed maintenance as decoupled FastAPI microservices responding to pub/sub events, enabling integration of new hardware models or external CMMS tools without touching core data flows.
• Security-first Mindset: Implemented multi-tenant JWT scopes early, avoiding costly retrofits as the client scaled to new regions.

Built a layered dashboard experience in ReactJS:

• At the top level, execs see site-wide utilization, ESG offsets, and cost savings.
• Mid-level lets engineers drill down to daily generation vs forecast curves and string-level inefficiencies.
• Bottom-level views show inverter anomaly timelines, with one-click drill into raw sensor streams.

Custom download modules generate board-ready PDFs on energy & ESG metrics — frequently used by CFOs to secure green investments. Included a maintenance heatmap calendar highlighting predicted interventions, helping maintenance heads plan workforce weeks ahead.

• Improved solar utilization by 20% via proactive load and maintenance strategies.
• Cut manual inspections by 40%, reallocating engineers to strategic expansions.
• Enabled the client to secure two rounds of green funding with traceable ESG data.