New blog: Do you know what your Claude Code is doing? (Part 2) Read the post →

FinOps, Observability and Optimization for your AI and Data

Autonomous agents that find wasted spend, slow queries and bad data across your LLMs, data warehouses, and then fix them.

Get a Demo Try the Sandbox →

See every AI dollar, and how much of it you can recover

AI spend · last 30 days

Live

$2.5M

+12% vs prior 30 days

Forecast · next 30d

$2.1M

Spend per call

$29.93

Active users

620

Requests

83K

Failure rate

12.0%

Top movers

Recoverable spend

See what every agent costs, and which ones are growing fastest

Spend by agent · last 30 days

$2.6M

Share of spend by agent

Daily spend by agent

Token throughput by agent

See how fast each agent responds, and how often it reuses cached work

Agent performance · last 30 days

Avg latency

Cache hit rate

30K

agent requests

2 agents

Latency and cache hit rate per agent

Write better and efficient code

Claude Code terminal

Repeated work this week

$412

went to 5 tasks Claude Code redid in every session. The top three:

Resolving Google Docs comments one at a time

$176

Writing a new log parser for every investigation

$79

Checking cron health by hand

$68

Fewer turns

Lower bills

Faster sessions

See every warehouse right-sized, hour by hour

Warehouse performance report · Adhoc_Workload_WH

$44,369.52

annualized savings

54% reduction

Avg query execution

17 s

was 25.13 s

Avg queuing time

1 s

was 3.71 s

Queries

54,504

was 37,203

Credits used weekly

264.42

at $3 per credit

Dynamic warehouse sizing heatmap: hours optimized vs no change across the week

Gartner Cool Vendor 2025 badge

Recognized as a

Gartner® Cool Vendor

2025

Trusted by global brands

Built for every AI model, Agent and every database you run

One platform. Reduces cost, improves performance and quality of your stack.

AI FinOps widget showing spend by model

AI FinOps & Observability

For every LLM and Agent

Observability and FinOps for LLMs and agents, with user, agent and model attribution across OpenAI, Anthropic and Google.

Granular

Cost attribution

Full

Visibility per user, agent and model

Zero

code changes to set up

Explore AI FinOps →

AI DBA widget showing fixes applied to Snowflake, Databricks and BigQuery

AI DBA

For every database you manage

Tunes warehouses, fixes slow queries and reclaims idle spend across Snowflake, Databricks, BigQuery and Redshift.

30-70%

Lower data platform spend

10x

Operational efficiency

5 min

To first results

Explore AI DBA →

Maximize the ROI of your AI, agents and data

The problems teams bring to us, and what changes once Revefi is running.

You are experiencing

Unknown Budget

Uncontrolled token usage.

Track Adoption

Trust issues with AI output.

Unpredictable Performance

Revefi impact

AI Observability

Full Visibility

AI Metering

Cost Control

AI Quality

Verified Truth

Works across

and Your model

You are experiencing

Blown Budget

Unpredictable compute costs.

Poor Data Quality

Inaccurate data & lack of trust.

Fire-Fighting

Constant operational issues.

Revefi impact

Reduce Data Cost

by 30–70%

Increase Efficiency

by 10×

Get Results

in 5 mins

Works across

It doesn't just flag the problem. It opens the ticket and the PR.

Revefi's AI agent and MCP server work inside Claude Code. Here it found an over-provisioned Databricks job, filed the Jira issue and opened a pull request with the fix.

Graviton3 nodes instead of m5d.large

Photon vectorized runtime

All checks passed, ready to merge

See it on your stack

Claude Code · ~/revefi/shasta-databricks

> Find the biggest Databricks saving and fix it

Revefi MCP: ETLMedJob is over-provisioned on m5d.large with the STANDARD runtime

✓ Created Jira issue TEST-416

✓ Opened PR #1 on revefi/shasta-databricks

TEST-416

IN PROGRESS

Optimize ETLMedJob cluster configuration with Photon runtime

ASSIGNEE CLAUDE (BOT) · LABELS COST-OPTIMIZATION, DATABRICKS

dbc_bundle/resources/etl_med_job.yml

+6 -2

23

spot_bid_price_percent: 100

24

-

node_type_id: m5d.large

24

+

node_type_id: m7gd.large

25

+

driver_node_type_id: m7gd.large

31

+

spark_conf:

33

+

spark.databricks.cluster.profile: singleNode

32

-

runtime_engine: STANDARD

36

+

runtime_engine: PHOTON

✓ All checks passed

Merge pull request

◀ ▶ ↻

Projects / TEST / TEST-416

TEST-416

Optimize ETLMedJob cluster configuration with Photon runtime

Description

Updated ETLMedJob cluster configuration based on Revefi's cluster optimization experiment recommendations.

Configuration Changes:

Node type: m5d.large → m7gd.large (Graviton3)
Runtime: STANDARD → PHOTON (vectorized engine)
Added driver_node_type_id and spark_conf for single-node

Benefits:

Better performance and cost efficiency for Auto Loader streaming workloads.

Activity

C

Claude created this issue

Just now

C

Claude changed status to In Progress

Just now

[TEST-416][Part 1] Optimize ETLMedJob cluster config with Photon runtime and m7gd.large #1

● OpenClaude wants to merge into main from kr_optimize-etlmed-cluster

Conversation

Files changed

Checks

▾ dbc_bundle/resources/etl_med_job.yml

+6 −2 · YAML

⋯ 20 unchanged lines

21 first_on_demand: 1

22 zone_id: us-west-2d

23 spot_bid_price_percent: 100

24 − node_type_id: m5d.large

24 + node_type_id: m7gd.large

25 + driver_node_type_id: m7gd.large

26 cluster_log_conf:

27 volumes:

28 destination: /Volumes/revefi/default/logs

29 spark_env_vars:

30 PYSPARK_PYTHON: /databricks/python3/bin/python3

31 + spark_conf:

32 + spark.master: local[*, 4]

33 + spark.databricks.cluster.profile: singleNode

34 enable_elastic_disk: false

35 data_security_mode: DATA_SECURITY_MODE_DEDICATED

32 − runtime_engine: STANDARD

36 + runtime_engine: PHOTON

37 kind: CLASSIC_PREVIEW

38 is_single_node: true

⋯ 4 unchanged lines

✓ All checks passed

Customer stories

1 / 6

Fortune 500 Insurer

Fortune 1000 Insurtech

Healthcare provider

Fortune 200 · Databricks

Fortune 50 · BigQuery

A Fortune 500 Insurer using Snowflake.

711

Snowflake Warehouses

50%

Spend Reduced

<48h

Results in <48 hours

Snowflake spend dropping 50% within 48 hours of connecting Revefi

A Fortune 1000 Insurtech company.

“With Revefi, the value of FinOps and Data Observability became clear within just a few months. They identified savings, improved quality and more than paid for themselves in no time.”

Louis DiModugno

Global CDO

Fortune 1000 Company

Read Case Study ↗

5 minutes to first insight, up to 30-70% lower cloud data costs, 665K+ automated monitors, 100% projected annual ROI

Healthcare provider using Snowflake.

27%

Less Snowflake spend

5×

Beyond the 5–6% expected

3 Weeks

To deliver, start to finish

Bar chart: 5-6% expected savings versus 27% actual savings

Fortune 200 Enterprise using Databricks.

~$1M

Saved during the free trial

25%

Of production covered so far

>$3M

Projected across all production

Bar chart: $1.0M saved in the free trial versus $3.0M projected savings

A Fortune 50 service provider, the largest BigQuery customer in the U.S.

100% Unified

All Data-SLAs and RCA

3x Increased

Interoperability between systems & tools

4 Platforms

In one interface: BigQuery, Looker, Composer & Gemini

Revefi connected to Gemini, Google BigQuery, Looker and Google Cloud Composer

Maximizing ROI of AI, F200 Enterprise.

8x

increase in AI spend in less than 4 months

25K+

users and increasing

200M+

requests monthly and increasing

10K+

AI services and increasing

How Revefi works: automated monitoring, alerting, recommendations and one-click actions

What makes Revefi different?

Most tools stop at the alert. Revefi runs the whole loop for you across AI and data: it monitors, alerts, recommends and acts, then keeps tuning on its own.

Self-tuning

Adjusts to workload shifts on its own.

Self-driving

Applies the fix, not just the alert.

Agentic AI

An agent on every model and warehouse.

Zero-touch

Live in 5 minutes, no code changes.

Revefi closed loop: monitor, alert, recommend and act across FinOps, observability and operations for AI and data

See what Revefi could save you.

See Revefi on your own stack in a demo, or estimate your savings first with our calculators for Snowflake, Databricks, Google BigQuery and AI.

Cost calculator

AI & LLMs Snowflake Databricks BigQuery

Monthly LLM and AI spend

$120,000

Estimated savings

$24K to $48K

a month

See it in a demo