Private Research AI

A privately hosted, Gemini-style web app, built in Regina, Saskatchewan, for a Canadian trader who organizes company filings into folders and asks questions against one folder at a time.

Client: Canadian client (name withheld) · Industry: Independent Trading & Financial Research

Services

Custom AI Web Application
Document Question Answering
Conversation History
Folder-Based Access Control
Private Cloud Hosting

Technology stack

Python
LangChain
Pinecone
OpenAI API
AWS

Project overview

A Canadian trader needed a research assistant that could work through large sets of company filings the way a product like Gemini answers a question — but only from documents he chose, and only inside a network he controlled. A consumer subscription could chat in general terms. It could not keep quarterly reports organized by market, sector, and company, answer from one folder, save that research, or later be handed to a team with limited access. Peradeo Technologies Inc., based in Regina, Saskatchewan, Canada, built a custom AI web application in Python, hosted on AWS elastic servers and reached over a private network. The client is Canadian, and his name is withheld.

The challenge

The research library was too large, and too sensitive, for a shared chat subscription:

• Company and industry reports had to live in a folder tree he defined — for example USA / Technology / NVDA — with quarterly filings and other relevant reports uploaded into that folder
• A question had to be answered from that folder, not from the entire library and not from the open web
• Every question and answer needed to be saved so he could revisit a thread, edit a question, or ask a new one beside it
• Research was going to move from one person to a team, and each team could see only its own folders
• The application had to stay on a private network
• He wanted a product he owned: something he could change and fine-tune without moving up a subscription tier

Our solution

We delivered a private research application with the question-and-answer feel of a hosted chat product, built around his documents:

• Folder libraries: He creates folders such as USA, Technology, and NVDA, then uploads quarterly reports and related documents into the folder he wants to study
• Folder-scoped answers: LangChain retrieves relevant passages from a Pinecone vector index for that folder, and the OpenAI API answers from those passages
• Saved research threads: Questions and answers are stored. He can reopen a thread, edit a previous question, or ask a new one
• Team access: When research moved to a team, each team was granted access to its own folders and not to the rest of the library
• Private AWS hosting: The Python application runs on AWS elastic servers and is reached over a private network, so filings stay off the public internet
• An owned system: Prompts, retrieval, access rules, and the model can be changed or fine-tuned on his terms, without a higher consumer-subscription fee

Related services from the Regina team: AI development in Regina, Python development, and AWS cloud hosting.

Results & impact

Regina, SK — built by Peradeo Technologies Inc. in Regina, Saskatchewan, Canada
Private — reached over a private network, outside a public chat subscription
Folder-Scoped — answers drawn from the selected library, such as USA / Technology / NVDA
Saved Threads — questions and answers kept, editable, and available to revisit
Team Folders — each research team sees only the folders it is granted
Owned Stack — Python, LangChain, Pinecone, and OpenAI on AWS, adaptable without a subscription tier
Canadian Client — built for a Canadian client whose name is withheld

Frequently asked questions

Who built this private research AI?

Peradeo Technologies Inc., a software company in Regina, Saskatchewan, Canada, built the web application for a Canadian client. The client’s name is withheld.

Where was the project delivered from?

The work was delivered by Peradeo’s team in Regina, Saskatchewan, for a Canadian client. The application runs on AWS elastic servers and is reached over a private network.

What does the application do?

A Canadian trader creates folders such as USA / Technology / NVDA, uploads quarterly and related reports, and asks questions against that folder. LangChain retrieves passages from a Pinecone index and the OpenAI API answers from those passages. Questions and answers are saved and can be edited or revisited. Each research team can open only its own folders.

Is this a public ChatGPT or Gemini subscription?

No. It is a custom web application the client owns. Prompts, retrieval, access rules, and the model can be changed or fine-tuned without moving to a higher consumer-subscription tier.

Related services

More case studies

Have a similar project? Contact us · Email hello@peradeo.com · Phone (Canada) +1 (306) 216-9934