The AI Engineer: How Savekar CMS Predicts Faults Before They Happen
Savekar insights
The AI Engineer: How Savekar CMS Predicts Faults Before They Happen
Learn how Savekar’s AI-driven Charging Management System predicts EV charger failures before they occur, helping businesses maintain uptime, reduce maintenance costs, and improve charging reliability.
13 Mar 202615 min read
The overall public EV charging infrastructure in India has expanded. It has increased from 5,151 charging stations in December 2022 to 25,202 in December 2024 and then to 29,277 in August 2025, thanks to the efforts of FAME and PM E-DRIVE. However, with the increasing infrastructure, it has been found that the lack of reliability and poor O&M are turning out to be major constraints, as opposed to the availability of charging points.
Savekar’s cloud-based Charging Management System (CMS) is developed to cater to the needs of the Indian market, which includes real-time analysis, remote control, and round-the-clock support to help property owners and managers achieve high uptime and, at the same time, earn revenue through UPI payments. This article explains how an "AI engineer" in Savekar’s CMS can track chargers all day long, predict failures before they occur, and use maintenance as a differentiator for B2B customers in India.
Why maintenance anxiety is real in India
The Government of India recognises that "obsolete technology, underutilization, and poor O&M" are major factors for the unreliability of public charging points in several states. Case studies in Delhi and other cities have found that equipment is missing or not working, there is no power supply, and there is no enforcement of EV-exclusive parking areas, which are factors that lower the availability of the charging stations and investor confidence.
However, the number of infrastructure developments is growing very rapidly, as evidenced by the official figures from the Ministry of Power (MoP), which show that the number of public EV charging stations rose from 5,151 in December 2022 to 11,903 in December 2023, then to 25,202 in December 2024, and finally to 29,277 in August 2025.
AIFault Diagnosis
India’s EVCS growth at a glance
The graph above illustrates the MoP/BEE figures (as reported through PIB and research institutions) to demonstrate the rapid expansion of public EV charging points.
For B2B customers like hotels, residential societies, fueling stations, offices, and fleet management, this expansion gives rise to two issues:
Whether the charging points will be functional when customers arrive
How much time and money will be wasted waiting for the service team to fix the issues
An intelligent CMS that functions like an “AI engineer” tackles both issues.
What predictive maintenance is
The process of predictive maintenance forecasts when a component will fail in the near future using data and algorithms. This guarantees that rather than waiting for the failure to occur, maintenance can be done just in time. This entails monitoring voltage, current, temperature, connector usage, communication errors, and session failure events in order to spot warning signs when it comes to EV charging.
Research on smart EV charging stations shows that, through the use of sensors, IoT, and machine learning algorithms, it is possible to reduce unplanned downtime by predicting failures in chargers, cables, power electronics, and safety systems. Also, experiences with predictive maintenance software for EV charging stations worldwide show that monitoring and maintaining the stations 24/7, instead of waiting for a failure to happen, leads to increased uptime and reduced costs for emergency repairs.
Policy context: India’s emphasis on accessible, interoperable, and reliable charging
The Ministry of Power has released the 2024 Guidelines for Installation and Operation of Electric Vehicle Charging Infrastructure. These guidelines are meant to ensure compatibility and reliability of charging configurations in India. The new guidelines are based on the existing guidelines and aim to ensure the communication of interfaces, open data reporting, and the development of public charging infrastructure through initiatives such as PM E-DRIVE.
Under these guidelines, the Bureau of Energy Efficiency (BEE) has been identified as the Central Nodal Agency for public EV charging, and it maintains a national database of charging points and manages the EV Yatra website, which helps users locate charging points. According to independent research, in the future, charging policies must not only support capex but also promote high uptime through effective O&M, data sharing, and standardised monitoring and verification of charger performance.
In the context of B2B customers, this policy shift will imply that reliability, data sharing, and standards will increasingly become considerations in the selection of charging solutions for commercial premises, campuses, and depots.
Savekar CMS: built for Indian property owners
Savekar CMS is designed with Indian property owners in mind. It is described as “India’s most affordable CMS with WhatsApp integration,” and its purpose is to assist property owners in monetising EV charging without the need for heavy software or complicated apps. The main features include direct UPI QR code payments (no app download necessary), real-time analytics, cloud-based dashboard, remote management of chargers, and 24/7 support. The cost begins at ₹1,999 per charger per year.
The solutions are designed for hotels, housing societies, small commercial properties, and other locations that wish to monetise EV charging. The solutions include a profit-sharing plan (no upfront capex; Savekar manages the operation and maintenance) or a full control plan (property owner purchases the chargers and retains 100% of the revenue). These solutions are designed to meet a common problem for Indian B2B businesses: how to begin and scale EV charging with low risk, easy payments, and reliable operational support.
The "AI engineer" within Savekar CMS
Though Savekar’s official website showcases real-time analytics, remote control, and 24/7 support, these features are only natural for an AI-assisted maintenance feature that works like a virtual engineer monitoring all chargers. This "AI engineer" can be imagined as a set of models and algorithms running on top of the CMS data pipeline, learning from charger data continuously to predict failures before they affect users.
At a macro-level, an AI-enhanced CMS system for EV charging in India would:
Gather data from charging stations, including charging start and end times, energy consumption, charging connector temperature, grid voltage fluctuations, error messages, network status, and payment success or failure.
While adhering to data privacy and security standards, analyze this data in the cloud to identify each charging station, location, and supplier.
Use algorithms that point to abnormal trends (such as rising errors, longer handshakes, and payment failures) as warning signals.
Send maintenance alerts or warnings on WhatsApp or dashboard notifications before a complete shutdown, so that ground-level personnel can take corrective action during off-peak hours.
In other words, instead of getting a complaint from a customer about "charger number 3 not working," the customer would get a warning about "charger number 3 possibly going down soon" and be suggested an advanced check.
Typical EV charger failure modes in India
Analysis of smart EV charging stations and practical experience with public charging networks indicate that there are a number of typical failure points that, if left unchecked, can lead to unforeseen periods of downtime.
Failure mode
Practical example in India
What the AI engineer watches
Hardware degradation
Connectors, cables, or relays wear out faster in high‑heat, dusty, or coastal environments
Rising connector temperature, abnormal current draw, and frequent plug‑in failures
Grid instability
Voltage dips or surges from the distribution network cause charger resets or protective shutdowns
Patterns of undervoltage/overvoltage, frequent restarts linked to grid events
Communication faults
OCPP/OCPP-like communication drops between the charger and CMS, leading to failed authorisation or remote control
Session failures at handshake stage, repeated network error codes
Payment and user interface errors
QR codes not scanning, UPI failures, or user confusion leading to abandoned sessions
High rate of partial sessions, payment declines, or aborted transactions
Poor onsite practices
ICEing (non‑EV cars blocking bays), power switches turned off, or cables missing
Long periods of zero utilisation despite local EV demand, irregular offline/online cycles
The classic maintenance approach will identify these issues only after customers have complained or usage rates have fallen significantly. An AI-powered CMS would be able to turn each of these into a signal that can prompt preventive maintenance.sss
How the AI engineer prevents faults before they happen
1. Early warning analytics for charger condition
Predictive maintenance analytics for EV charging analyzes continuous sensor inputs (current, voltage, temperature, and power quality) to forecast when components are trending outside normal parameters. By learning what constitutes a healthy charger for each model and location, predictive models can point out anomalies days or weeks before they become apparent.
On Savekar’s CMS, this type of intelligence can be layered behind the existing real-time analytics dashboard, working in the background to assign a score to each charger’s health and trigger internal notifications when risk reaches a certain level. This means that for B2B customers, the only information presented is a simple status (e.g., green, amber, red) and recommended course of action.
2. Intelligent prioritisation of maintenance visits
With the expected explosive expansion of the Indian EVCS network, it is no longer feasible or economical to dispatch maintenance personnel in a reactive manner for each reported fault. By using predictive analytics, the most critical sites can be prioritised, ensuring that maintenance resources are first allocated to those EVCSs that have a high likelihood of failing in the near term or are of high revenue significance.
For instance, a hotel on the highway with a high usage rate and increasing error messages would be prioritised over a residential charger with a low usage rate and functioning correctly.
3. Remote fixes and configuration optimisation
Most problems can be fixed remotely by firmware upgrades, configuration adjustments, or simple resets, especially when communication between the charger and the CMS is strong and standards-compliant. Savekar’s CMS system already has the capability for remote management, including cloud-based monitoring and control of chargers, which allows for remote fixes without the need to send a technician every time.
The AI layer can also learn which error combinations are amenable to remote fixes and make recommendations or automate safe fixes, further minimising downtime and on-site expenses. This builds a knowledge base of typical problems and their known fixes over time, specific to Indian hardware, grid, and user behaviour.
4. Feedback into vendor and asset decisions
Since Savekar’s system integrates chargers from various vendors and locations, anonymised performance insights can help identify which chargers are robust in a particular set of Indian conditions and which ones tend to need frequent fixes. Predictive analysis of this data can then assist B2B customers in selecting the optimal mix and replacement timing of chargers, focusing on minimising lifecycle costs instead of initial capital expenditures.
This is particularly important as government programs like PM E-DRIVE and state EV policies encourage the massive deployment of both slow and fast charging stations, where the long-term operational and maintenance costs can easily surpass the upfront costs if the reliability is low.
Business value for Indian B2B customers
International best practices, including those in geographies with uptime mandates, indicate that predictive maintenance and 24/7 monitoring can help ensure charger availability of 97 per cent or higher and lower the number of emergency repairs. Indian studies emphasise the need for enhanced O&M and uptime-related incentives to enhance customer experience and protect public investment in chargers.
Savekar already provides high uptime support through its on-ground service team and 24/7 support services as part of its CMS solution. With AI-based maintenance, this adds up to substantial business value for B2B customers:
Uptime: Successful charging sessions and happy EV owners result in direct customer satisfaction and retention for hotels, malls, and offices.
Total maintenance cost: With fewer emergency repairs, predictable O&M budgets become possible instead of surprise costs.
Stable and growing revenue: Direct UPI transactions, subscription systems, and profit-sharing deals ensure that every successful charging session increases site revenue immediately.
Compliance advantage: With government guidelines becoming increasingly tough on reliability and data, sites with intelligent CMS and strong logs are better placed to face audits or rewards for site uptime.
How Savekar supports different B2B archetypes
Savekar’s business models are designed to suit the needs of Indian property owners who may not want to become full-time charge point operators. Two typical archetypes are used to demonstrate how an AI-powered CMS can alleviate maintenance concerns for property owners:
Zero-capex property owners (Profit-sharing model)
Savekar installs its 3.3 kW AC chargers, operates, and maintains them, and splits the profits with the property owner.
The AI engineer is always keeping tabs on the health and usage of the chargers; property owners are simply concerned with revenue and basic status information.
In both scenarios, the message to B2B clients is clear: they do not need to build an in-house data science or engineering team to benefit from predictive maintenance. The "AI engineer" is already embedded in the CMS layer and provided as part of the service.
Practical questions B2B buyers should ask
For the purpose of building trust and mutual understanding, the following set of questions can be asked by Indian B2B buyers interested in Savekar CMS (or any EV charging CMS) to ensure mutual understanding of expectations:
What are the current levels of uptime that you maintain in your network, and how are these measured (per connector, per site, per month)?
How does your CMS identify warning signs of a failure, and what kind of problems are normally fixed remotely versus site visits?
How soon are problems assessed after being identified by the system or reported by users?
How are software updates and configuration changes implemented without interrupting operations, particularly in high-demand locations?
What types of dashboards and reporting tools are available for facility managers to view the status of each charger for health, usage, earnings, and maintenance status?
How is data protection handled, and how does the system comply with Indian standards for interoperability, communication standards, and data sharing with public platforms such as EV Yatra, if applicable?
A clear and data-driven answer to these questions, coupled with AI-driven monitoring, helps to reassure that the CMS is more than a billing system—it is a long-term operational partner.
Conclusion: From maintenance risk to competitive moat
India's EV charging infrastructure has advanced to the point where it is now a business and policy issue rather than merely a technical one. This suggests that rather than isolated and underutilised charging infrastructure, government regulations and initiatives are calling for more interconnected, interoperable, and well-maintained infrastructure.
By considering Savekar CMS as an "AI engineer" who is constantly monitoring chargers, predicting failures, and advising on appropriate action, B2B customers in India can turn maintenance risk into a differentiator. This means chargers that just work, payments that just happen, and an investment that continues to pay back dividends for many years, rather than being a recurring source of trouble.
Postscript: technical notes for interested readers
A. Data sources for predictive maintenance
The following data sources should be made available by smart EV charging stations that support predictive maintenance via proprietary APIs or communication protocols like OCPP:
Electrical: power factor, energy consumption, harmonics, events on the power grid, voltage, current, and each phase.
Thermal: temperatures of internal modules, cables and connectors, and, if available, ambient air.
Event logs: The event logs include firmware version numbers, warnings, error messages, start and stop events, and resets.
Communication metrics: Communication performance includes the charger and CMS's signal strength, packet loss, latency, and reasons for disconnections.
Transactional data: session duration, energy consumed in kWh, payment status, and user identifiers or token types (if privacy-compliant).
In the Indian scenario, these data sources should be handled with proper cybersecurity and, if applicable, in compliance with guidelines on data accessibility and interoperability issued or recommended by nodal organisations such as BEE and MoP.
B. Model families employed as an "AI engineer"
Predictive maintenance of EV charging stations often involves a combination of statistical and machine learning models:
Anomaly detection models (such as autoencoders or clustering models) are trained on historical "normal" data to identify anomalies in real-time.
Survival and reliability models to estimate the remaining useful life of components based on stress variables such as high temperature, frequent fast charging, or low-quality grid.
Time-series forecasting models (such as ARIMA, gradient boosting, or neural networks) to forecast future error rates or probabilities of downtime based on past trends.
Rule-based expert systems constructed from field engineering knowledge that implement known patterns such as "if the connector temperature exceeds the threshold more than N times in a week, inspect".
These models can be implemented in the cloud together with the CMS, with periodic re-training as more Indian data becomes available, enabling the "AI engineer" to adjust to local hardware, climate, and usage patterns.
C. Architecture alignment with Indian guidelines
From an architecture standpoint, an AI-facilitated CMS in India would have a tiered architecture:
Device layer: EV charging stations and energy meters talking to each other through standardised protocols (e.g., OCPP) over secure communication channels.
Connectivity and ingestion layer: Edge gateways or cloud interfaces accepting data, checking formats, and buffering data during network downtime.
Core CMS layer: Services for user management, tariff management, access control, payments (UPI, cards, and subscriptions), and basic monitoring.
AI/analytics layer: Data lake, feature engineering pipelines, model training and scoring engines producing health scores, predictions, and recommendations.
Integration and reporting layer: Operator dashboards, APIs for third-party apps, and data export capabilities that can be aligned with government websites such as EV Yatra or other reporting interfaces, as may be required in the future.
Organising such a system with interoperability and data sharing in mind puts B2B customers in a position to take advantage of future incentive programs that pay for uptime and open performance data, while ensuring that today's customers get reliable and predictable charging at their facilities.
