ENERGY SYSTEMS
Unlock subsurface potential.
Connect geological and operational data to evaluate resources, compare development scenarios, and support long-term performance.
GeothermalOil & gasCarbon storage
SCIENCE GROUNDED. AI ENABLED.
Intelligence
beneath the
surface.
A deeper understanding of our planet.
A clearer path for energy and the environment.
Scientific AI that connects the two.
THE LIVING SUBSURFACEExplore what lies beneath
X / −32° Y / 28°
DRAG TO ROTATE · CONCEPTUAL MODEL, NOT FIELD DATAAI FOR ENERGY & ENVIRONMENTSCROLL TO EXPLORE ↓
BUILT FOR COMPLEX SYSTEMS
LESS GUESSWORK. ＋ MORE UNDERSTANDING.
From scattered data to connected decisions.WHAT WE DO
Complex earth.
Clearer possibilities.
We bring geoscience, physical models, and AI together
to help teams understand uncertainty —
and decide what comes next.
ENVIRONMENTAL INTELLIGENCE
Understand risk. Plan ahead.
Model groundwater, contaminant transport, and subsurface response to inform site assessment, remediation, and environmental planning.
Water systemsRemediationSeismic risk
SCIENTIFIC AI
Put science inside the model.
Build physics-informed workflows that connect data preparation, model calibration, and scenario analysis — with people in control.
Physics-informed AIAgentic workflows
REAL-WORLD APPLICATIONS
One planet.
Many ways to make a difference.
Explore where deeper understanding
can move your project forward.
SELECT A SYSTEM
EXPLORE THE POSSIBILITIES
GEOTHERMAL ENERGY
Find the heat.
Understand the opportunity.
Bring subsurface observations and thermal models together to evaluate geothermal resources, plan development, and explore production scenarios.
- Resource evaluation and site screening
- Reservoir modeling and uncertainty analysis
- Development and operating scenarios
DATA FOUNDATION
Temperature–depth profiles, injection and production histories, pressure-transient tests, tracer responses, well geometry, and geological structure.
ANALYTICAL METHODS
Thermal–hydraulic coupling, reservoir history matching, fracture-connectivity alternatives, and sensitivity to permeability and well spacing.
DECISION OUTPUTS
Pressure interference, thermal-breakthrough scenarios, production-temperature envelopes, and priorities for further reservoir testing.
THE SCIENCE BEHIND THE DECISION
Domain knowledge.
Computational depth.
Not every problem needs a neural network.
Start with the physics, interrogate the data,
and choose the method the decision demands.
MODEL ARCHITECTURE
OBSERVEField & operational dataLogs · cores · sensors · spatial layers
↓REPRESENTConceptual & numerical modelsGeometry · physics · boundary conditions
↓LEARN & TESTCalibration + scientific AIInverse modeling · surrogates · validation
↓DECIDEScenarios with uncertaintyTradeoffs · evidence · decision thresholds
Model architecture is selected for each project. Numerical simulation, machine learning, and expert interpretation play different roles.
RESERVOIR ENGINEERINGCoupled subsurface processes
＋
Coupled subsurface processes
Represent fluid flow, heat transfer, and pressure response in porous or fractured formations. Begin with thermal–hydraulic (TH) coupling; introduce mechanical or chemical coupling where stress change, fracture behavior, or water–rock interaction materially affects the decision.
- Inputs
- Well trajectories, lithology, pressure–temperature logs, permeability, porosity, fluid properties, and operating history.
- Methods
- Finite-volume / finite-element simulation, fracture representation, history matching, and parameter sensitivity studies.
- Outputs
- Pressure and temperature fields, well interference, thermal-breakthrough scenarios, and operating envelopes.
The appropriate level of coupling depends on data support and project purpose. A more complex model is not automatically a better model.
HYDROGEOLOGY & TRANSPORTGroundwater and contaminant fate
＋
Groundwater and contaminant fate
Develop a conceptual site model (CSM) before simulating groundwater flow and solute transport. Distinguish advection and dispersion from sorption, decay, and biogeochemical transformation rather than fitting a single attenuation rate to every process.
- Inputs
- Hydraulic heads, pumping and recharge, well screens, analytical results with detection limits, and hydrostratigraphic units.
- Methods
- Transient flow, advection–dispersion modeling, particle tracking, and process-appropriate reaction models.
- Outputs
- Flow paths, receptor concentration scenarios, capture-zone assessments, and monitoring-network priorities.
A solute-transport model is not automatically a general geochemical reaction solver. Chemistry and density effects need an appropriate formulation.
Technical background: USGS MODFLOW 6 ↗CARBON STORAGEInjectivity, migration, and containment
＋
Injectivity, migration, and containment
Evaluate CO₂–brine multiphase flow alongside storage-complex geology. Analyze the pressure footprint separately from the CO₂ plume, and test how permeability architecture, relative permeability, and capillary pressure influence migration.
- Inputs
- Storage and seal geometry, petrophysical logs, core measurements, pressure data, fluid composition, and injection schedules.
- Methods
- Multiphase reservoir simulation, alternative geological realizations, pressure-response analysis, and monitoring scenario design.
- Outputs
- Plume and pressure envelopes, injectivity constraints, uncertainty registers, and technical monitoring rationale.
Model results support technical review; they do not constitute a permit or a guarantee of storage performance.
Technical background: EPA Class VI guidance ↗PHYSICS-INFORMED MACHINE LEARNINGLearning within physical constraints
＋
Learning within physical constraints
Physics-informed neural networks (PINNs) can combine observations with differential-equation residuals and initial or boundary conditions. Reduced-order models and Gaussian-process surrogates offer alternative ways to accelerate repeated evaluations.
L = λdLdata + λpLphysics + λbLboundary
Weights, variable scaling, and sampling strategy affect training. Evaluate performance against numerical and statistical baselines on withheld conditions; low training loss alone does not establish predictive accuracy.
- Deliverables
- A documented training domain, benchmark results, model limitations, and a fallback route when inputs fall outside the validated range.
INVERSE METHODS & UNCERTAINTYQuantify what the data cannot resolve
＋
Quantify what the data cannot resolve
Calibration narrows plausible parameter values; it does not eliminate uncertainty. Separate measurement error, parameter uncertainty, and alternative conceptual models when evaluating the reliability of a prediction.
- Methods
- Regularized parameter estimation, ensemble or Bayesian updating, Monte Carlo / Latin hypercube sampling, and global sensitivity analysis.
- Review criteria
- Residual structure, parameter identifiability, predictive interval coverage, and sensitivity of the decision to alternative assumptions.
- Outputs
- Scenario ensembles, uncertainty intervals, parameter-influence rankings, and priorities for additional data collection.
SCIENTIFIC SOFTWARE & AGENTIC AIAutomate the workflow, retain oversight
＋
Automate the workflow, retain oversight
Use task-oriented agents to assist with data validation, run preparation, experiment comparison, and report assembly. Keep authoritative calculations in versioned scientific tools, not in free-form language-model responses.
- Controls to scope
- Typed tool interfaces, explicit permissions, approved data boundaries, run logs, review gates, and bounded retry policies.
- Delivery options
- Batch pipelines, analyst workspaces, API-backed applications, or deployment within a customer-controlled environment.
- Human review
- Model assumptions, operational changes, and final technical conclusions remain subject to designated expert approval.
Integration and security requirements are agreed during discovery; no certification or production integration is implied.
INDUSTRY CASE STUDIES
Where the science
meets the field.
Technical perspectives on documented industry projects.
Explore the evidence, the methods,
and the questions they raise for the next project.
3 project analyses
Flow.
Connection.↗
UTAH FORGE · UNIVERSITY OF UTAH
Reading an engineered geothermal reservoir.
How circulation testing, fiber-optic sensing, and production measurements inform reservoir interpretation.
EGSInjectivityFiber optics
Read the technical analysis ↗Monitor.
Verify.↗
ILLINOIS BASIN–DECATUR · MGSC & PARTNERS
Storage performance needs more than a plume map.
A closer look at subsurface characterization and integrated monitoring in a deep saline storage project.
CO₂–brineMVAContainment
Read the technical analysis ↗Understand.
Respond.↗
CAPE COD RESEARCH SITE · USGS
Why contaminant behavior changes along a flow path.
Field tracer experiments reveal the importance of transport, sorption, and changing geochemical conditions.
Tracer testsRedoxTransport
Read the technical analysis ↗MODEL ASSURANCE & TECHNICAL DELIVERY
A result is only useful
if it can be examined.
Define the acceptance criteria before modeling begins.
Deliver the reasoning alongside the result.
Data lineage
Track source files, units, coordinate reference systems, timestamps, quality flags, and preprocessing decisions. Preserve original observations separately from interpreted values.
Numerical checks
Inspect convergence, conservation balances, mesh and time-step sensitivity, and boundary-condition behavior before interpreting an attractive visualization.
Independent evaluation
Hold back wells, locations, or time periods where appropriate. Test residual bias and interval coverage, and document conditions where performance deteriorates.
Reproducible handover
Package configurations, data dictionaries, run manifests, model versions, review records, and operating instructions within the agreed licensing and data boundaries.
A TECHNICAL PACKAGE, NOT JUST A DASHBOARD
What a scoped engagement can deliver
The statement of work defines which artifacts are included and who owns each review gate.
Define your deliverables ↗01Data & conceptual modelSource register / CSM / assumptions
02Model & evaluation recordConfiguration / calibration / validation
03Decision & uncertainty reportScenarios / sensitivities / limitations
04Handover & implementation planRunbook / versioning / next measurements
FROM DATA TO DECISIONS
Not a black box.
A clearer line of sight.
A prediction is only part of the answer.
Understand where it comes from, where its limits are,
and what it means for your next decision.
Connect the data.
Unify geological, monitoring, and operational inputs while retaining their origin, scale, and quality.
Ground it in physics.
Combine scientific knowledge with machine learning. Calibrate, validate, and quantify uncertainty.
Explore what changes.
Compare scenarios and identify the variables that matter to performance, risk, and resilience.
Make it actionable.
Translate findings into traceable reports and repeatable workflows, with expert review at key decisions.
THE PRINCIPLES BEHIND THE PROCESS
Explainable · Traceable · Reproducible
ABOUT ENVITRACE
Earth science.
Built for real decisions.
Envitracer LLC operates under the envitrace brand, focusing on scientific AI software and technical services for energy and environmental systems.
Our work connects geoscience, physics-based modeling, and data analysis to support resource evaluation, environmental risk assessment, and project planning. Engagements are shaped around the question, the available evidence, and the decision a team needs to make.
LEGAL ENTITYEnvitracer LLC
BUSINESS ADDRESS548 Juneau Seawalk
Juneau, Alaska 99802
United States
Juneau, Alaska 99802
United States
FOCUSScientific AI · Energy · Environment
Pilot projects PROVE THE FIT
Start with a focused use case. Define the data, evaluation criteria, and deliverables before committing to a larger program.
Custom AI & deployment BUILD FOR YOUR TEAM
Shape models, analytical tools, and interfaces around your data and workflows. Choose deployment options around your security and collaboration needs.
Technical consulting FIND A CLEAR PATH
Get support with data readiness, modeling strategy, interpretation, and implementation planning.
Research partnerships EXPLORE TOGETHER
Collaborate on scientific modeling, method validation, and applied research with clearly defined outputs and intellectual property arrangements.
BEFORE THE FIRST MODEL RUN
Questions worth
asking early.
Technical clarity starts before procurement.
Here is how to frame the conversation.
What data is needed to get started?
Start with the decision to be supported, the site or system boundary, available observations, and a data inventory. Useful inputs may include well logs, geological interpretations, pressure and temperature records, water levels, chemistry, and operating schedules. Include units, coordinate systems, sampling methods, and quality flags. A discovery review identifies gaps before a full model is commissioned.
When is physics-informed AI appropriate?
When there is a well-defined physical formulation and a clear reason to combine it with learning. PINNs, surrogate models, and conventional numerical solvers solve different problems. Method selection should consider data coverage, required fidelity, repeated-run cost, and validation evidence. AI should earn its place by improving an agreed benchmark, not by being included in the project title.
How do you evaluate predictive performance?
Agree on decision-specific metrics before calibration. These may include bias and error against independent observations, prediction-interval coverage, conservation residuals, or sensitivity of a scenario ranking. Spatially or temporally correlated observations require appropriate validation splits. A strong fit to calibration data is not the same as successful prediction.
Can the workflow use existing models and software?
Begin with a review of the existing model, solver version, mesh, assumptions, output conventions, and licensing. Integration may use structured exports, batch interfaces, or documented APIs. A proof-of-compatibility task should precede commitments to production integration. Proprietary solver access and redistribution rights remain subject to their own licenses.
How are confidential data and deployment handled?
Define data ownership, permitted processing locations, retention, access roles, and third-party services in the project agreement. Customer-controlled or isolated deployment can be evaluated during scoping. Security controls and any compliance requirements must be verified for the actual implementation rather than inferred from a website description.
What does a pilot include, and how long does it take?
A pilot should specify a bounded question, usable input data, a baseline method, evaluation criteria, and tangible outputs. The schedule depends on data readiness, solver complexity, computing requirements, and review cycles. The first deliverable is an agreed scope and evidence plan; a fixed timeline should follow that review, not precede it.
LET’S LOOK DEEPER
Big questions.
Start a conversation.
Tell us what you’re working on.
We’ll start with your challenge, not a sales pitch.
COMPANYEnvitracer LLC
LOCATION548 Juneau Seawalk, Juneau, Alaska 99802, United States
WEBSITEenvitracer.com ↗
CONTACT