Company Research Assistant

Read 500 pages every day. That’s how knowledge works. It builds up, like compound interest.

The Company Research Assistant, an AI-powered tool, is designed with the goal to help you, an active investor, study a company more efficiently and build up your knowledge faster.

Get Started

To work with the assistant, you just need 3 steps: import your data, defining analyzing logic in modules, and output.

Use Cases

Example 1: Ask questions from imported documents.

          from Tesla import 2023_PR.pdf,2022_PR.pdf,2021_PR.pdf

          ask "How many cars did Tesla sell in 2023?"
          ask "How many cars did Tesla sell in 2022?"
          ask "How many cars did Tesla sell in 2021?"
        

You can upload various data sources such as text files, PDFs or Data API results. You can also create a folder and import all files under the folder.

The assistant will only perform tasks within the scope of your data. This eliminates the hallucinations of just using a Large Language Model.

You can import multiple times from different sources, and use imformation via variables and tags from previous function calls.

Example 2: Factsheet comp across companies

          //factsheet comps across companies

          import BOA_2023_10K.pdf
          import JPM_2023_10K.pdf
          
          [BOA]
          ai (get_content BOA_2023_10K.pdf,"get the 2023 BOA income statements and balance sheet in table")
          [/BOA]
          
          [JPM]
          ai (get_content JPM_2023_10K.pdf,"get the 2023 JP Morgan income statements and balance sheet in table")
          [/JPM]
          
          output 'Do a side-by-side comparision of BOA financials in BOA tag and JPM financials in JPM tag.'
        

Example 3: Summarize a file

          from examples import chat.pdf

          get_content chat.pdf

          output "Summarize all the questions in Korean." 
        

Thanks to the Large Language Model under the hood, the assistant is cross languages, meaning you can ask in any natrual langues and it can output in any langues.

Example 4: Extract your highlights and comments of PDF

A common company research scenario is that you make highlights and comments while reading along the 10Ks and 10Qs. The assistant can help you extract these hightlights and reformat them as your study notes.

          //Get the highlights from a pdf, and perform a reformat.
          import HES_2023_10K.pdf
          
          get_highlights HES_2023_10K.pdf
          
          get_notes HES_2023_10K.pdf
          
          output 'summarize my highlights and notes in bullet points. output them in Xmind format'
        

Here is what it looks like after we import the output to Xmind.

Example 5: Call Data APIs and analyze the results

            //Google News monitoring
              $result1 = googlenews '腾讯'

              $result2 = googlenews '阿里巴巴'

              ai ($result1, "列出最负面的3条新闻，按时间排序")
              ai ($result2, "列出最负面的3条新闻，按时间排序")

              ai ($result1, "列出最正面的3条新闻，按时间排序")
              ai ($result2, "列出最正面的3条新闻，按时间排序")
          

The assistant can easily integrate Data Feed APIs. Currently it supports Yahoo Finance, Google News and Wikipedia.

Example 6: Analyze and calculate

          analyze "Define P/S ratio as its market value divided by its revenue. \
          What's NVIDIA's P/S ratio?"
        

The assistant will try to solve the problem by first breaking down this complex question to a sequence of InvRet function code. Make sure its syntax correct so we can directly run it to get the real answer.

Here is what the assistant generates from above.

          $marketCap = yfinance NVDA,marketCap
          import NVDA_2023_10K.pdf
          [revenue]
          ask 'Get NVIDIA 2024 total revenue.'  
          [/revenue]
          $ps_ratio = ai ($marketCap+$revenue, 'Calculate Apple P/S ratio by dividing the market cap by the 2022 total revenue. Output the P/S ratio as a number.')
          output $ps_ratio
        

The more specific the question you want to analyze, the eaiser it is for the assistant to generate an actional solution. The assistant can't help much if you ask a very general question such as what are the stocks to buy.