I am a portfolio manager of a small/mid-cap equity fund. This is definitely cool and clearly took a lot of time and thought!
If this is a side project, that's amazing and I love it. If you did want to take it further though, I would though encourage you to spend some time interviewing potential users and thinking through how you are creating value.
As a fundamentals investor, to win in stockpicking (especially against quant traders), you need to get really deep into the details. As mentioned there are lots of tools that summarize earnings calls, but more importantly, I don't think summarizing earnings calls or releases are super useful for investors.
There are basically two ways to build useful software for investors. One way is to create summaries / off-the-shelf analyses that tell investors if a company is good or bad. Right now as-is, that is sort of where this product is. The problem is, an earnings call summary is insufficient. Yes, AAPL revenue may be up 4.0%, but the fact that you figured that out in 15 seconds with an LLM is not a differentiator since everyone else can figure it out in 45 seconds by just doing mental math. And, what did _you_ expect it to be? Very different trade answer if you thought it should be down 10% or up 40% and it in reality hit 4.0%.
I think generally this is also tough because the people really good about building recommenders (black box or otherwise) just work at hedge funds and make a ton of money for themselves / their clients. It is almost always going to be more profitable to do this, in no small part because once you reveal your secret sauce to the market, it gets priced in.
There's also the issue of user personas. There is no one generic investment strategy. A "good" investment is very different for a long-only vs. a long-short vs. a macro fund vs. an activist fund (and yes these are not mutually exclusive at all).
So, I think in order for this to become truly useful you need to make it more of an interactive tool to allow people to run their own analysis (vs having a "one size fits all" analysis). A few ways you could do this:
1) investments are done behind an investment thesis. Can you TRACK theses over time? For example, a merger arb thesis might be that a big M&A deal is on track (or not on track) to close. Can you track my THESIS, which almost becomes a detailed LLM prompt. E.g., "I think that Company X's acquisition of Company Y will close at the current offered price of $Z by date A, and key milestones are (1) regulatory approval on or around date B, (2) stock price converging toward the offered price of $Z as date A approaches, (3) positive sentiment on earnings calls about the acquisition by both Company X and Company Y"; let me know if earnings releases or calls or 8-Ks should make me update my thesis ("8-K came out that said that the target date of A has slipped by 90 days"). Or, "Company X said they bought Company Y and at the time expected revenues of $Z in three years; what compound quarterly growth rate does this imply" and then track that every quarter to see if they are on track or not
2) Can you help me track a specific question through each quarter—almost like automated BI / reported? Bloomberg/CapIQ/etc. will do this for GAAP metrics like revenue or COGS but often companies will report bespoke metrics. Can you help me track WeWork's community-adjusted EBITDA (lol) every quarter? Can you help me track business unit non-GAAP EBITDA with an addback for taxes every quarter? These are things that won't show up in your GAAP 3-statement model but for _individual_ companies may be super important
3) similarly can you help me run comparisons across companies? E.g., can I compare Vornado's total revenue (almost all office) to Bridge's office segment's revenue to SL Green's total revenue (also almost all office) and every quarter see how their relative revenue (market share) is changing? What about their occupancy and rent levels?
4) generic RAG/natural language search (when did SmartRent talk about their acquisition of Planon), although as noted, many competitors are doing this
All of the above are things that analysts do in real life and basically just track manually in Excel.
Again don't mean to be discouraging—I posted this in part because I think there is a ton of value here to (a) having a clean pipeline and (b) having a basic working knowledge of applying LLMs in a useful way. If helpful I'm happy to discuss more.