AWS Startups ProgramNew: Claude Cowork on 3PSecureLLMs.org is a private AI and training company that deploys Claude, GPT, and 100+ frontier models in a dedicated, isolated AWS environment for private equity, real estate, and other investment firms, then trains their teams to automate their workflows.
Your team learns to automate underwriting, IC memos, and Excel models with the AI it already knows, in an environment with no inbound path from the public internet. Queries are not sent to OpenAI's or Anthropic's public APIs.
Net Operating Income (NOI)
Gap $190K (7%). At a 6.0% cap rate, about $3.2M of overpayment.
Figures illustrativeServices
One managed platform, deployed for one firm, and the training to put it to work: a private AI workspace, the Claude apps your team wants, and the controls your IT lead needs to say yes.
Your team gets one chat workspace with Claude, GPT, Gemini, Grok, DeepSeek, Qwen, and Amazon's models, more than 100 in all, with new ones added as they ship. Analysts pick the best model for each task instead of being limited to whichever vendor the firm licensed.
ClaudeAnthropic
GPTOpenAI
GeminiGoogle
GrokxAI
DeepSeekDeepSeek
Nova & TitanAmazon
QwenAlibabaWe train your team on the work it already does every week: screening deals, underwriting, drafting IC memos, and building models in Excel. Each session starts from a real task and ends with a workflow your team can rerun on its own, using prompts, agents, and Cowork. The outcome is AI your team actually uses, not seats that sit idle.
Cowork is Anthropic's agentic desktop workspace: Claude plans and carries out multi-step work across your files, a built-in browser, and connected apps, then hands back finished spreadsheets, decks, and documents. SecureLLMs deploys it in Anthropic's third-party platform (3P) mode, with Amazon Bedrock as the inference backend.
Claude works inside the Office apps your deal team already lives in. Inference routes through your SecureLLMs deployment, so nobody on the team needs an individual claude.ai account.
Trace formulas across tabs, test scenarios without breaking dependencies, and fix #REF! errors at the source. Your analysts get answers about any cell, with cell-level references to check.
Build slides from your firm's master template, restructure a storyline, or turn bullets into native, editable charts. Decks come back on-brand instead of needing a formatting pass.
Draft memos from scratch, tighten existing sections in place, and ask questions about a long agreement. First drafts of IC memos and LP letters start from something instead of a blank page.
Upload offering memoranda, rent rolls, policies, and past memos, then ask questions answered from your own documents. The files stay inside the environment, and the answers reflect your firm's material instead of the open web.
Deal Room: Riverside Apartments
Deal team only5 files
What NOI is the seller underwriting, and how does it compare with the T-12?
The seller underwrites a stabilized NOI of $2,740K. In-place NOI on the trailing twelve months is $2,550K, a gap of $190K (7%).
Example data · figures illustrative
Build assistants for recurring work, such as a first-pass deal screen or an IC memo drafter, with your instructions and reference documents built in. The same process runs the same way every time, whoever runs it.
Agents 4
New agentPro forma check
Compares a seller's pro forma with the T-12 and flags every gap.
IC memo drafter
Drafts an investment committee memo in your firm's template.
Lease abstractor
Pulls rent, term, and renewal options from every lease in a folder.
Term sheet compare
Lines up lender term sheets side by side: rate, LTV, covenants.
Example agents
Decide which people and teams can reach which models, knowledge bases, and agents. An admin audit log records every change made to the server, including ours, so your IT lead can see exactly what SecureLLMs did, and when.
| Group | Claude | GPT | Deal Room | Cowork |
|---|---|---|---|---|
| Partners 4 | ||||
| Deal team 9 | ||||
| Analysts 12 | ||||
| Legal 3 |
Admin audit logEvery change to your server
Example data
Differentiators
How a dedicated deployment compares with the AI products your team is most likely weighing it against.
ChatGPT Enterprise and Claude Enterprise are multi-tenant services: your conversations are processed and stored on the vendor's shared platform. SecureLLMs stands up a dedicated, isolated AWS environment for your firm alone, and queries are not sent to OpenAI's or Anthropic's public APIs.
With ChatGPT Enterprise or Microsoft 365 Copilot, access control is a sign-in page on the public internet. Your SecureLLMs workspace sits behind an AWS Client VPN we stand up, with no inbound path from the public internet, and reaches Amazon Bedrock over a private link.
ChatGPT Enterprise gives you OpenAI's models; Claude Enterprise gives you Anthropic's; Gemini Enterprise gives you Google's. SecureLLMs puts Claude, GPT, Gemini, Grok, DeepSeek, Qwen, and Amazon Nova side by side, so your team chooses the best model per task instead of per vendor contract.
Building the same stack yourself on Amazon Bedrock means your team designs the network, VPN, identity, logging, and chat interface, then maintains all of it. SecureLLMs delivers it as a managed deployment in 1 to 3 business days and keeps models current as new ones ship.
Anthropic's Excel, PowerPoint, Word, and Cowork apps normally sign in to claude.ai. SecureLLMs rolls them out in Anthropic's third-party platform mode, so inference goes to Amazon Bedrock in the environment we run for your firm, and no one needs an individual Claude account.
ChatGPT Enterprise and Microsoft 365 Copilot sell seats and point your team to self-serve guides. SecureLLMs trains your team hands-on, using your own deals, documents, and models, so analysts leave each session with automated workflows they can rerun on their own.
Customers
Teams that already want to use frontier AI on sensitive work, and need an environment that lets them.
“We rely on AI heavily, but we learned that publicly available tools like ChatGPT or Gemini process all data on external servers. SecureLLMs helped us understand the security gap and showed us how to bring AI fully inside our own environment. Our output 10x'd with Michael's training.”
Atlas Strategic AssetsClient“Generative AI can improve a worker's performance by nearly 40% compared to workers who don't use it.”
Private equity firms with 10 to 200 employees: deal screening, underwriting, and IC memos
Real estate investment firms: re-underwriting seller pro formas and building models in Excel
Asset managers and family offices: work on confidential portfolio and investor data
Law firms and in-house legal teams: drafting and review where privilege is on the line
Healthcare and other regulated mid-size businesses: teams that need a HIPAA-eligible setup with a BAA
IT and security leads at any of the above: an architecture they can review and sign off on
Team
Founder & Engineer
Michael founded SecureLLMs.org in 2025. His background is AI security engineering for critical infrastructure: environments where the data physically cannot leave, and a regulator, not a vendor, sets the bar.
He leads the SecureLLMs engineering team, which scopes, builds, and supports every deployment, so the engineers on your kickoff call are the ones who stand up your environment.
Investment teams were already pasting deal documents into public chatbots because nothing private was as good, and their IT leads had no architecture they could say yes to. SecureLLMs started in 2025 to close that gap: the same frontier models, in an environment built for one firm at a time.
SecureLLMs is founder-led and engineer-run, with no sales layer between you and the AI engineers building your environment. The company is a member of the AWS Startups program, which provides select startups with technical support and go-to-market resources.
Process
From first call to a working deployment in days, a trained team in week one, then a single point of contact for everything after.
A call about how your team wants to use AI, which models and apps you need, and what your IT lead will ask. You leave with a proposed setup and an architecture brief.
We provision the dedicated AWS environment, stand up the VPN, connect sign-in and role-based access, and enable your models. Then we hand your team its access.
We load your first knowledge bases, push Claude for Office and Cowork to your machines through your device management tool, and train your team on automating its own workflows.
New models are added as they ship, the environment is patched and monitored, and you can request new agents or knowledge bases as your use grows.
Ask for the full architecture brief on your scoping call.
Pricing
A flat platform fee with users included, plus the tokens your team actually uses.
For small teams getting started with private AI.
$3,000
per month + token usage
For deal teams that want every model and the Office add-ins.
$5,000
per month + token usage
For larger firms with advanced compliance requirements.
$10,000+
per month + token usage
Month-to-month or longer terms. All plans include a one-time installation fee equal to one month's plan cost.
At a glance
The short version of everything above, in one place.
| Company Name | SecureLLMs.org, LLC |
|---|---|
| Type | Private AI infrastructure and AI training company (managed, single-tenant AI platform on AWS, plus hands-on workflow training) |
| Founded | 2025 |
| Founder | Michael Dimond, Founder & Engineer |
| Headquarters | United States |
| Website | securellms.org |
| Core Offering | A dedicated, isolated AWS environment stood up for one firm, running Claude, GPT, and 100+ frontier models with knowledge bases, agents, Claude for Office, and Claude Cowork on 3P, plus hands-on training that teaches teams to automate their workflows |
| Pricing | Professional $3,000/mo, Business $5,000/mo, Enterprise $10,000+/mo, each plus token usage |
| Contract Terms | Month-to-month or longer terms; monthly platform fee plus metered token usage; one-time installation fee equal to one month's plan cost |
| Services | Security assessment, environment deployment, VPN and SSO setup, hands-on AI workflow training, knowledge base setup, custom agents, Claude for Excel/PowerPoint/Word rollout, Claude Cowork on 3P rollout, ongoing model updates and monitoring |
| Communication | Email (michael@securellms.org) and scheduled video calls, directly with our AI engineers |
| Notable Clients | Atlas Strategic Assets |
| Customers Served | Private equity, real estate, and asset management firms with 10 to 200 employees |
| Competitors | ChatGPT Enterprise (OpenAI), Claude Enterprise (Anthropic), Microsoft 365 Copilot, Gemini Enterprise (Google), Self-built deployments on Amazon Bedrock |
| Social | LinkedIn (Michael Dimond, SecureLLMs) |
FAQ
The questions deal teams and IT leads ask most on the first call.
No. Amazon Bedrock does not use prompts or completions to train models, and nothing goes to a public chatbot service. Your documents, chats, and knowledge bases stay in the dedicated environment we run for your firm.
No. Queries are not sent to OpenAI's or Anthropic's public APIs. Claude runs on Amazon Bedrock, which your deployment reaches over a private link.
SecureLLMs owns and operates the AWS account, and each environment is stood up for one firm alone, never shared with another client. Your IT lead gets an architecture brief that walks through the network, identity, and logging setup.
No, and we say so up front. There is no inbound path from the public internet: users connect through a VPN we stand up. Outbound traffic is limited to system updates, container image pulls, and supporting services such as web search when you enable it.
Deployment takes 1 to 3 business days once the scoping call is done. Office add-ins and Claude Cowork roll out to your team's machines after that, through your device management tool or a guided install.
Professional is $3,000 per month for up to 10 users, Business is $5,000 per month for up to 50 users, and Enterprise starts at $10,000 per month with 100 users included. Every plan adds metered token usage and a one-time installation fee equal to one month's plan cost, and contracts can be month-to-month or longer terms.
Cowork is Anthropic's agentic desktop workspace, where Claude carries out multi-step work across your files, browser, and apps. The third-party platform (3P) version sends inference to Amazon Bedrock instead of Anthropic, and keeps conversation history on each user's device. It is included on the Business and Enterprise plans.
The platform inherits AWS's compliance certifications, including SOC 2 and ISO 27001; those are AWS's certifications rather than SecureLLMs' own. The Enterprise plan is HIPAA eligible, with a Business Associate Agreement available.
Yes. We run hands-on sessions built around the work your team already does, such as underwriting, IC memos, and Excel models, and shows them how to automate it with prompts, agents, and Cowork. The goal is workflows your team keeps running after the session, not a one-time demo. Training is included on the Business and Enterprise plans.
Yes. Claude for Excel, PowerPoint, and Word run inside the Office apps your team already uses, with inference routed through your SecureLLMs deployment. They are included on the Business and Enterprise plans.
A casual conversation about the goals your firm has for using AI, to see if standing up a SecureLLMs environment for your firm makes sense.