Career Sage
Effortlessly filter cross-platform career information and improve competitiveness with tailored AI mentorship.
Created on 9th July 2023
•
Career Sage
Effortlessly filter cross-platform career information and improve competitiveness with tailored AI mentorship.
The problem Career Sage solves
High Cost of Trial and Error: College students have very limited years to learn and try before making final career choices; students often struggle to ascertain if their efforts truly guide them towards their career objectives.
Overwhelmed Information Filtering: In the sea of cross-platform career information, students often feel overwhelmed to filter valuable ones and even miss crucial deadlines for scholarships or career opportunities.
Lack of Prompt Tailored Mentorship: 90% of students need tailored career mentorship, but students find it challenging to secure mentors who can provide practical tailored guidance to improve their competitiveness at any time.
Challenges we ran into
The major challenges we ran into are the unreliable outputs from GPT, especially the GPT3.5 models we currently are using to control the budget. During our LLM-chain to plan courses for students, we need to access our private database stored in Milvus, and we put this a s tool for GPT to use or query. During the query process, GPT also needs to give the keywords in a specific format. However, it might give nothing or give additional useless explanations that crash the process later. So we have to write some redundant code to deal with these issues. Due to the time limit, we couldn't try more, but we are considering combining GPT4 and GPT3.5 to let GPT4 make the high-level choices and let GPT3.5 combine information and give answers (use lots of tokens).
Tracks Applied (3)