Batteries bring the flexibility modern grids depend on. We ensure they scale profitably.

Physics-informed machine learning models

Grounded by battery domain experts

Built for gigawatt scale assets

For The BESS Ecosystem

  • Developers
  • Owners
  • Operators
  • Financers
  • Insurers
Teammates AI Harness demo screenshot
odessa-bess-200mw · q2 outlook · post rtc+b
Will we hit Q2 proforma?
Objective: hold Q2 EBITDA at proforma, availability above 97%
CEO 8m ago
@reliability How certain are we about meeting Q2 proforma post RTC+B?
@reliability AI 5m ago
+0s pulled SCADA telemetry, last 90 days
+1m cross-checked NOAA weather forecast, Apr 28 to 30
+3m modeled C7 thermal derate against ERCOT load
Two swing days at risk: Apr 28 and Apr 30. ERCOT peak load coincides with C7 thermal derate trend. Availability projects to 95.8% versus 97% guarantee. LD exposure ~$38k. Routing to @engineering with the maintenance brief.
Sources: SCADA, ERCOT, NOAA → Amperical physics based ML model
@engineering AI just now
Acknowledged. Pulling C7 HVAC service forward to Apr 26, blocking dispatch in that window. Handing to @qse to re-bid Apr 28 and 30 conservatively.
Sources: OEM schedule, SCADA, dispatch logs → Amperical physics based ML model

The Problem

The asset doesn't read the proforma

Markets, weather, and cells write the actual P&L.

Where problems

Location and product mix

  • Origination and interconnection dictate which markets the asset can reach
  • Winning product mix shifts as grids evolve
  • Regulations and incentives shape the proforma

What problems

Specs and design

  • OEM, sizing, augmentation cadence locked at design
  • Specs lock in the trajectory for life
  • Hard to reverse, choices compound for 20 years

How problems

Operations

  • Unplanned maintenance erodes availability
  • Value stream mix shifts realtime across energy, ancillary, capacity
  • Shifting supply-demand dynamics compress spreads

Product

Intelligence for BESS Proforma Reconciliation

Simulate pre-COD, then track proforma vs actual for the full lifetime.

Origination

  • Grid congestion screening
  • Value stack per market
  • Incentives and tariffs mapping

Simulations

  • Pre-bid P&L scenarios
  • True cost discovery
  • Sensitivity analysis

Tracking

  • Live proforma vs actual
  • Augmentation and availability tracking
  • Early red flag alerts

How It Works

Physics based empirical models

Modeling Approach

Physics-informed machine learning trained on real operations.

Spreadsheets and MATLAB lack machine learning. Generic AI lacks domain. We sit where deep BESS physics meets modern machine learning.

ML + AI → BESS expertise → Spreadsheets MATLAB Energy analytics Generic AI Amperical
Neural network architecture: customer telemetry, location, design and specs flow through physics-informed ML layers to produce actionable intelligence

Training Pipeline

Pre-trained, fine-tuned, expert-validated

Pre-training
Lab cycling data, simulated conditions
→
Fine-tuning
Historical telemetry, per-project
→
Expert Validation
Domain experts + physics guardrails

Agentic AI Harness

Specialized teammates with cited reasoning

Reasoning trace
100% traceability and grounding
→
Hand-off across the team
BESS and market tools built-in
→
Finding posted with sources
Shareable and actionable results

Team

Combining deep energy + AI expertise

Rachana Vidhi, PhD

Co-founder & President

  • 10 US patents in batteries and renewables enabling 1,000MW+ of BESS deployments
  • PhD Chemical Engineering (USF), Master's in Management (Harvard), IIT Kharagpur
  • Former Director, NextEra Energy
Rachana in Florida Inventors Hall of Fame →

Indresh Kumar

Co-founder & CEO

  • Pioneered BESS analytics at Ion Energy: Amazon backed, scaled to 600MWh+ AUM
  • Developed battery fleet operations AI/ML models at VoltUp
  • Technical co-founder: AI, ML and data models
Indresh on ~600MWh BESS analytics in California →