Intelligence Applied

Operational AI for the
Indian EV battery chain.

Field data in. One decision per battery per week — for every stakeholder who depends on it.

Talk to us →
For the first time —
Battery OEM
Sees how their design performs in Indian field conditions.
↳ Not the lab.
NBFC / Lender
Knows the health of the asset they lent against. Before default.
↳ Not the invoice. Not the court.
Fleet Operator
Knows which vehicle stops — six weeks before it stops.
↳ Not after.
Battery OEM
Your lab test results cannot show what 42°C Indian field conditions do to your cells.

Pack3401 degrading at −0.0077/week vs Pack3001 at −0.0013/week — a 6× gap invisible in the lab, confirmed across 27 batteries, 3 cities, 13 months. Attribution: operator charging 34%, commissioning 28%, IoT gaps 22%, calendar 16%. Warranty settled in 90 seconds. ₹11.2K liability — not ₹28.8K.

Fleet Operator
Which batteries need action this week — and which of those have no symptom yet?

14 batteries flagged for unnecessary replacement. All 14 were NOT replaced. ₹5.6L avoided in one quarter. Knowing when not to act is worth more than knowing when to act. One action per battery per week. Replace, swap charger, monitor, nothing.

NBFC / Lender
You are pricing a loan against the one asset component you cannot assess.

Default from the asset, not from credit. ₹2.8Cr collateral gap on a ₹12Cr book — decomposed in 12 minutes: 38% operator behaviour, 29% OEM design, 21% use case stress, 12% calendar aging. 8–12 weeks before first missed EMI. Not after the default. Before.

Vehicle OEM · Regulator
The range figure on your vehicles is not the range in the Indian field.

Cell imbalance alert firmware is disabled in most deployed Indian fleets to reduce nuisance alarms. The safety signal is turned off. enerlytik detects cell imbalance independently — without the BMS alert. 4.4km range overstated every trip. +5.5% SoC firmware bias confirmed across 179 batteries.

The condition
The chain exists.
The problem
The connections didn't.
enerlytik
Now they do.
Intelligence Applied →
The gap that costs the most
BMS Reports
94.2%
State of Health
NO ALERT · NO ACTION
FIELD TRUTH
enerlytik Reads
63%
True electrochemical health
REPLACE IN 6 WEEKS · ₹28.8K CLAIM INCOMING
The BMS was designed to protect the cell — not to inform the business.
LFP's flat OCV plateau makes BMS SoH fundamentally unreliable. This is physics, not firmware.
500+
batteries scored weekly
5+ cities · 3 operating climates
97M+
CAN telemetry rows
204 features per battery per week
52
ML models in production
22 intelligence chains
5
stakeholder surfaces live
OEM · Fleet · NBFC · Driver · Regulator
13+
months continuous field data
Multiple load · temp · driving profiles
0
false negatives
Across all model changes
PRODUCTION GRADE · AWS DEPLOYED · WEEKLY SCORING · API-FIRST ARCHITECTURE · 5 STAKEHOLDER SURFACES LIVE · 52 ML MODELS IN PRODUCTION · 0 FALSE NEGATIVES · 97M+ CAN TELEMETRY ROWS · DPIIT RECOGNISED · PHYSICS BEFORE ML ·

PELT detected the regime change.
11 weeks before the floor breach.

The same methodology quantitative traders use to detect structural breaks in equity markets — applied to your battery fleet. The signal existed for months. The BMS never saw it.

BMS reported SoH
Field truth (DEKF + physics)
Conformal band (82.4% coverage)
100% 90% 80% 70% 60% 50% Wk 1 Wk 10 Wk 20 Wk 30 Wk 34 Wk 40 Wk 50 FLOOR PELT · REGIME CHANGE WEEK 34 · −0.0077/wk FORECAST CUSUM DRIFT ALERT 14 weeks below baseline BMS: 94% — NO ALERT FIELD: 81% — DECLINING FLOOR BREACH 82.4% CONFIDENCE
BatteryBAT · LFP · A (anonymised field asset)
MethodPELT changepoint · CUSUM drift · DEKF state estimation · Conformal prediction
Lead time11 weeks before floor breach · BMS alert: never fired
Warranty attribution · before the claim arrives

5-factor attribution. 90 seconds. Physics-grounded.

The same factor decomposition quant desks use for portfolio attribution — applied to battery warranty liability. SHAP R²=0.816. No dispute. No manual investigation.

BAT · LFP · CS-04 · Failed at month 14
VERDICT: PARTIAL · OEM ₹11.2K · OPERATOR ₹17.6K
Operator charging
56%
52 weeks irregular high-rate charging detected
OEM design
28%
Commissioning cell spread 195mV above fleet P75
IoT / commissioning gaps
22%
14 weeks partial telemetry — signal gaps noted
Calendar aging
16%
Expected for 14-month LFP under Indian climate
ModelSHAP decomposition · XGBoost ensemble · R²=0.816
vs manual2 hours vs 3 weeks. Same outcome. Evidence-backed.
Saving₹17.6K — OEM liability reduced from ₹28.8K to ₹11.2K
Live fleet intelligence

One decision per battery per week.
Automatically.

Scoring runs every week across the full fleet. PELT detects regime changes. CUSUM flags drift. Conformal bands set confidence. enerlystAI delivers the action.

FLEET INTELLIGENCE · WEEK 47 · LIVE SCORING
SCORING ACTIVE
Battery
SoH
Trend
Signal
Horizon
Action
MethodPELT · CUSUM · DEKF · Cox PH · Conformal bands · SHAP attribution
CadenceWeekly scoring · 204 features per battery · 52 ML models

If this is your problem,
let's talk.

We work with a small number of partners.
No long commitment. Response within 48 hours.

Gaurav Mehra
Co-Founder · Commercial & BD
gaurav@enerlytik.com linkedin →
Natraj Akella
Co-Founder · Product & Technology
natraj@enerlytik.com linkedin →
● DPIIT Recognised
● AWS Deployed
● 500+ batteries live
Direct email: gaurav@enerlytik.com