Apna: Optimizing upskilling and job opportunities for millions across India with Vertex AI
Since launching in 2019, Apna has grown into India's largest jobs and professional networking platform by deploying an AI algorithm to connect millions of workers to relevant job opportunities. Apna built a cloud-native marketplace and job-matching platform on Google Cloud, using Vertex AI, Google Kubernetes Engine, and BigQuery to support continual model improvement, large-scale analytics, and platform safety.
Reported outcomes
60%
quantified impactOther quantified impact
Strategic outcomes
Primary read
Use case focus
Showing 3 of 4
- 1AI matchmaking
- 2Content moderation
- 3MLOps
- Match millions of blue-collar and professional workers with relevant jobs and upskilling opportunities at scale.
- Continuously improve matching models as employer and candidate needs change.
- Detect and remove abusive or fraudulent content on a large community platform.
- Built a proprietary AI job-matching algorithm on Vertex AI.
- Used BigQuery pipelines to process up to 500 million user interactions per day for analytics and modeling.
- Used GKE to run cloud-native microservices and deploy features quickly.
- Used Vertex AI-driven ML models to detect abusive or fraudulent behavior through keyword detection and improve platform safety.
- Ran multiple AI experiments per day to fine-tune matching performance.
- Estimated 20% reduction in time to create AI models.
- Up to seven AI experiments per day enabled.
- Estimated up to 40% savings in DevOps time.
- Up to 60% of inappropriate content removed daily.
Architecture
Apna's platform uses Google Kubernetes Engine for cloud-native microservices, BigQuery for high-scale data pipelines and analytics, and Vertex AI for model development, deployment, and daily experimentation. Vertex AI also supports ML models used to detect abusive or fraudulent content.
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?