I really need to learn more about the implementation of workflows outside of the agents. This is great! Have you thought about creating a course once you have this process down?
Thank you, Joel! I’ve actually been thinking about this, and I see two levels of consideration.
On one hand, with the pace of AI application development, a dedicated research tool like this might already be on the horizon, GPT-o3 and Perplexity, for instance, seem to be heading in that direction.
On the other hand, I believe there will still be niche needs in certain domains that generalized tools just can’t fully address (like the example I shared from my own experience).
So I’m definitely keeping an eye out for opportunities. Do you have any thoughts on this?
I’ve done many courses, and with the tools out today, it’s not hard to create them and update them. Even if you put a short course with some walkthroughs and in three or fours months technology eliminates its need, people are usually at least 6 months behind and would keep using your course.
I’m hoping to put out a little open-source version or MVP when I get the chance.
In the meantime, have you tried using ChatGPT-4o? I’ve found it surprisingly good for searching materials online, almost like having an AI agent do the digging for you.
I was actually working with a startup that wanted to do something like this to provide CI to pharmcos as well as investment analysis to traders.
The idea was to rank biotech NDA candidates based on the probability of approval and odds of marketplace success and the magnitude thereof. The input was to be clinical literature, but other EA would also doubtless be useful.
That’s a lot of info! When it comes to calculating probabilities and potential to guide decisions, things get tricky...
From my outsider’s perspective:
People likely already have a gut sense of how certain drug candidates are ranking, they just really need solid evidence to back that up.
So the first layer of gathering data sources should be pretty straightforward. That part’s mostly APIs or scraping, no need to involve AI.
The next step gets harder: if there aren’t well-defined rules for what makes a candidate promising, then you probably need AI to study past approvals and establish those patterns. Once that baseline is set, the model can interpret new clinical literature more meaningfully.
For external analyses, my experience is some are really useful, some are pure noise. Building a rule to weigh them properly would really help.
And yeah making it to FDA approval pipeline makes it much easier to decide.
I’d love to know more. Not sure how applicable it would be to my area (literature and language teaching- college level) but I have to write courses and I think this may help!
I really need to learn more about the implementation of workflows outside of the agents. This is great! Have you thought about creating a course once you have this process down?
Thank you, Joel! I’ve actually been thinking about this, and I see two levels of consideration.
On one hand, with the pace of AI application development, a dedicated research tool like this might already be on the horizon, GPT-o3 and Perplexity, for instance, seem to be heading in that direction.
On the other hand, I believe there will still be niche needs in certain domains that generalized tools just can’t fully address (like the example I shared from my own experience).
So I’m definitely keeping an eye out for opportunities. Do you have any thoughts on this?
I’ve done many courses, and with the tools out today, it’s not hard to create them and update them. Even if you put a short course with some walkthroughs and in three or fours months technology eliminates its need, people are usually at least 6 months behind and would keep using your course.
Besides, good revenue stream in the meantime :)
Thanks, they are really good insights :)
Great example. I would like to know more on AI Agent like the example. I want to search for journal article for Literature Review using Agent.
Thanks, Johan! Glad you found it useful.
I’m hoping to put out a little open-source version or MVP when I get the chance.
In the meantime, have you tried using ChatGPT-4o? I’ve found it surprisingly good for searching materials online, almost like having an AI agent do the digging for you.
Yes, I have tried using ChatGPT-4o. What are the suggestion in using it for Literature Review. Thanks.
Great to example, would love to hear about how you created the agent flow (even though it ‘lost’ 🙃) :)
Oh haha, I'd be happy to! 😊
I was actually working with a startup that wanted to do something like this to provide CI to pharmcos as well as investment analysis to traders.
The idea was to rank biotech NDA candidates based on the probability of approval and odds of marketplace success and the magnitude thereof. The input was to be clinical literature, but other EA would also doubtless be useful.
This obviously gets easier the further a
Compound gets down the FDA approval pipeline.
That’s a lot of info! When it comes to calculating probabilities and potential to guide decisions, things get tricky...
From my outsider’s perspective:
People likely already have a gut sense of how certain drug candidates are ranking, they just really need solid evidence to back that up.
So the first layer of gathering data sources should be pretty straightforward. That part’s mostly APIs or scraping, no need to involve AI.
The next step gets harder: if there aren’t well-defined rules for what makes a candidate promising, then you probably need AI to study past approvals and establish those patterns. Once that baseline is set, the model can interpret new clinical literature more meaningfully.
For external analyses, my experience is some are really useful, some are pure noise. Building a rule to weigh them properly would really help.
And yeah making it to FDA approval pipeline makes it much easier to decide.
You’re awesome
Awww thank you Kyle!
I’d love to know more. Not sure how applicable it would be to my area (literature and language teaching- college level) but I have to write courses and I think this may help!
Hi Sir,
I am interested. Can share your AI prompts and AI agent?
Truly appreciate,
Paul Goh
Hi, i would love to get the full workflow