To connect more unemployed youth with jobs and allow job seekers to apply via mobile apps, Harambee moved to Google Cloud Platform, helping the nonprofit to expand its mission with big data solutions and innovative machine learning algorithms.
In Swahili, the word “Harambee” means “all pull together,” reflecting an African tradition of community self-help. That’s exactly what’s happening in South Africa, a country with a high unemployment rate and one of the highest youth unemployment rates in the world. Out of a population of 55 million, more than 6 million young people are effectively locked out of the job market, with no work history and limited access to education or training. Unemployment levels for the 18 to 35 age group exceed 50%.
Harambee is an aptly named not-for-profit organization working to solve this problem. It works with local partners and government to match high-potential youths with entry-level positions. These applicants are hungry for opportunity but lack the financial-servicess and networks needed to find jobs.
"Google support means a lot to us and gives us even more credibility with employers as we work to solve the problem of youth unemployment in South Africa. But this is a worldwide problem, and Google Cloud Platform gives us global reach and shows a commitment to support our future goals."
Harambee needed to enhance its digital services to support new mobile site initiatives and make strategic use of the data it collected from assessments conducted at multiple touch points along job candidates’ journeys. Although Harambee collected a huge pool of data from candidates, it was difficult to analyze without scaling up compute resources. After evaluating nearly every cloud service available in April 2017, Harmabee decided to move from on-premises data centers to Google Cloud Platform (GCP). Although technology was a major factor in the decision, the Google shared commitment to Harambee’s cause through Google.org and Google for Nonprofits made a difference.
Committed to results that can work at scale, Harambee hopes that its solutions can someday help disadvantaged youths around the world.
"Google Cloud Platform provides the capacity, capabilities, and partner network we need to expand our services and our mission."
“Google support means a lot to us and gives us even more credibility with employers as we work to solve the problem of youth unemployment in South Africa,” says Evan Jones, CIO at Harambee. “But this is a worldwide problem, and Google Cloud Platform gives us global reach and a commitment to support our future goals.”
Adds Nicola Galombik, co-founder of Harambee Youth Employment Accelerator: “We had reached a point where we had hit the limits of what we could learn and what we could do in real time with that data. Google Cloud Platform opens up an enormous sense of possibility.”
Unlike some other cloud providers, Google offers Points of Presence (PoPs) in South Africa, allowing Harambee to use direct peering for its network traffic. Services running on Google Cloud use the Google global network to reach a Google PoP as close to them as possible, minimizing latency. As a result, Harambee needs only to provide network routing and redundancy to the Edge Network — after that, delivery to the data centers is the responsibility of Google. This architecture reduces costs, allowing Harambee to pay only for local data delivery.
“Using Google Cloud Platform reduced our number of potential failure points, and that was a big consideration for us,” says Navid Erfani-Ghadimi, Enterprise Architect at Harambee. “Google Cloud Platform provides the capacity, capabilities, and partner network we need to expand our services and our mission.”
To deliver services at scale across Africa, Harambee requires maximum flexibility. Harambee worked with DotModus, a Google Cloud Partner with a specialization in data analytics and a commitment to helping nonprofits succeed, to build a cloud-native architecture that would provide the flexibility to meet future requirements. To develop and host its web and mobile apps, Harambee chose the App Engine flexible environment in which application instances run within Docker containers on Compute Engine virtual machines. Microservices run on Kubernetes Engine for fast deployment and optimal scalability, while data is stored on Cloud Storage and in MySQL databases on Cloud SQL. For enterprise-grade access control, Harambee uses Cloud Identity & Access Management.
“DotModus provided excellent support from development through production, helping us understand how our data flows and machine learning algorithms would work best on Google Cloud Platform,” says Navid. “We saved months of time.”
Harambee has interacted with over 1 million young people to collect what it believes to be the largest dataset on youth employment in South Africa. Over 400,000 of those young people have been invited to attend a face-to-face work seeker support session at one of Harambee’s centers, and 50,000 have found employment through Harambee’s interventions. Harambee is using the data it collects to advance its mission and make more job placements, partnering with industry analysts to understand labor market dynamics, and working with organizations such as the World Bank to conduct cutting-edge research.
For entry-level jobs, the cost and duration of a worker’s commute significantly impact the success of the job placement. Therefore, Harambee takes great care to identify and verify candidates’ address information. Because many candidates live in areas that lack formal house numbering and street names, this can be a difficult task.
To analyze large datasets and provide rich visualization capabilities, Harambee uses BigQuery and Google Data Studio. It also uses Google Maps Platform to help calculate job seekers’ distances from prospective employers and match them to locations they can easily get to. After feeding data from mobile sites and contact center systems into Google Maps API to estimate an accurate location for the job seeker, results are stored in BigQuery. When employment opportunities arise in a given area, Harambee runs algorithms that extract location coordinates from all nearby job candidates and calculate their commute variables for optimal placement.
“Aside from people, data is our greatest asset, and it can help us improve critical metrics such as time and cost of transportation, absenteeism, and time on the job,” says Navid. “Geocoding with Google Maps APIs helps us do that, and BigQuery helps us enhance that analysis.”
Adds Marzanne Collins, Technology Development at Harambee: “Google Cloud Platform is helping us be a change agent in our job market. We can gather data about these young people and empirically prove to employers that our metrics more accurately reflect potential job success and return on investment compared with the exclusionary criteria they have been using.”
Harambee is using ML to match candidates to opportunities using precise geographical attributes that take informal and multi-transport routes into account. All candidates and job opportunities are now processed and stored in BigQuery. Candidates are then indexed to the Apache Solr open-source search platform. Opportunities are pulled from Google BigQuery into PySpark, building training data on inferences based on past behavior to build a vector map model that predicts ease of transport as a scalar score for all nearby candidates. The viable candidate-opportunity scores are then pushed to Solr for easy search and retrieval, helping reduce transport cost and time.
Harambee is also using ML to predict preferential behavioral metrics, including which candidates are most likely to leave a job within the first year of employment. Although transportation plays a role, Harambee wanted to go deeper and try to promote candidates whose past behavior favored long-lasting relationships with employers. To accomplish this, Harambee worked with DotModus to build TensorFlow models trained via Cloud Dataflow and saved in Cloud Storage. Candidate data stored in Apache Solr is pushed to Cloud Dataflow via Cloud Pub/Sub, where behavioral metrics are predicted and pushed back to Solr.
A job candidate’s journey now starts by submitting an employment application on one of Harambee’s new mobile sites. As a result, job seekers no longer have to travel to Harambee’s offices or mobile outreach units to connect with employers. Although smartphone penetration is low among underserved populations in South Africa, most citizens have some type of cell phone, and Harambee makes it easy for them to view jobs and complete applications using any device.
“Having responsive self-service from any device is very important to the youths we serve, many of whom have difficulty affording the trip to one of our offices,” says Bryony Maxwell, Head of Marketing for Harambee. “That’s just one example of how we’re using Google Cloud Platform and our own technology to provide a better experience for job seekers.”
“We can use Google Cloud Platform to measure behaviors and nudge job seekers into the right pathway over time. A candidate may start off being a cook and end up becoming a coder — we’re opening up a world of possibilities for our youth.”
As Harambee deploys new services, it can scale faster and easier because there is no longer any lag time for the provisioning and delivery of infrastructure. Instead of waiting months for procurement, delivery, and integration, Harambee can spin up a new server cluster on GCP in minutes — and at a much lower cost.
“Compared to our legacy IT costs, we expect to spend 70% less with Google Cloud Platform,” says Evan. “That’s money we can use to develop additional resources and services to help disadvantaged young people.”
If Harambee can make headway against youth unemployment in a country as diverse and complex as South Africa, it has a good chance of making a future impact in other countries — including other African countries such as Rwanda — and potentially benefit millions of youths. With GCP, Harambee can expand its digital services anywhere in the world and have confidence that they will be performant, available, and highly secure.
“No matter what the job market, we can use Google Cloud Platform to measure behaviors and nudge job seekers into the right pathway over time,” says Yatin Nana, Product Manager at Harambee. “For example, a candidate may start off being a cook and end up becoming a coder — we’re opening up a world of possibilities for our youth.”
Harambee Youth Employment Accelerator participated in Data Solutions for Change, a Google data analytics program for nonprofits to achieve their missions at scale. As a participant, Harambee received Google Cloud credits, self-training resources, and enablement support.