10xhr.ai

Inspiration

Our technical team spends a huge amount of time whenever new developers are hired. We have HR colleagues who do initial screenings, but very often, candidates come through who lack the needed skills. This then leads to many interviews that we feel could have been avoided. Then, while building a competency prediction system for educational purposes, we realised that the same core concepts can be used to solve this HR problem as well.

What it does

AI that helps HR managers identify high-quality tech candidates through real-time speech analysis & guided questions.

Core features:

Turn job posting into a competency tree
Based on real-time speech, identify the candidate's level of knowledge in the relevant skills for the job
Suggest questions for the interviewer that would allow getting more relevant information from the candidate
After the interview, create a summary of the candidate's suitability for the job role.

How we built it

Client-side POC web app - using LangChain, Gemini 3-pro-preview, grok 4-fast, langchain, education-context-inspired custom prompts

4 main prompts: Job posting → competency map Transcript chunks → competency assessment Transcript+job requirements → question suggestions Competencies → candidate suitability summary

Challenges we ran into

Didn't have more time to implement the full feature list.

Accomplishments that we're proud of

Fully working web app POC. A system where you see relevant skills being updated in real-time

What we learned

For real-time competency evaluation, chunk size and chunk number being sent for competency evaluation are crucial settings to optimise. This creates a classic speed vs quality tradeoff scenario.

What's next for 10xhr.ai

Ability for agent to join Zoom Speaker identification (for in-person use cases) Distinguishing between low knowledge vs no data in a competency Detailed competency evidence review Integration to other use-cases, such as supporting mentoring in education