SGInnovate Summation Programme
Apprenticeship with High-Potential Deep Tech ProjectsGain Real-World Experience Working on AI, Robotics, Quantum Tech and More

Project 1:

Social Media Inference Through Deep Learning

At Affable, we profile millions of users every day and use Machine Learning to infer things like: what are their interests? How old are they? Where are they located? We use a host of signals to make such inferences, including but not limited to their profile pictures, instagram posts and stories, and captions.

This project involves conceptualising, training and deploying Machine Learning models that can infer even more characteristics about social media users. Some of the characteristics we want to extract are psychographics, education level, job function, and income level.

The main challenges of this project are two-fold: volume and speed. Affable tracks over 1.5 billion social media posts and 300 million social media users. Machine Learning at this scale is challenging, time-consuming, but also very rewarding. You will get a first-hand experience in building models that are not only accurate, but also fast and are production-grade (some of our models run inference on over 20 million images per day).

Project 1:

Insurance Machine Learning by Google AI

Through technology and seamless integration, Axinan creates high-frequency products with a low premium due to accurate risk assessment based on our real-time risk engine. When dealing with fast-changing market condition and scenarios, it is important to be able to quickly create tailor-made products that can handle complex risk and fraud conditions.

Our technology is selected by Google for the latest cohort of startups for its Launchpad Studio program. With Google technologies and support from Google AI experts, we develop insurance products driven by AI and Machine Learning technology.

In this project, you will be working with the engineers and data scientists to develop cutting-edge Machine Learning models, improve accuracy and implement models in real-life insurance products. This includes models that analyse millions of features and models that run locally on mobile phones.

Project 1:

Extraction of Textual Streaming Data

Streaming textual data coming from online applications like chatbots, question forms and emails has its specifications in comparison with structured or semi-structured documented text. As usual, the text is written in free format, with missed punctuations, broken structures and spelling mistakes. The goal of this project is to extract all personal information from free text.

Project 1:

Deep Learning on Echocardiograms

We are building a global data network housed in some of the world's pre-eminent imaging labs and academic institutions. This data serves as the foundation for Machine Learning and statistical analysis for disease prediction studies. We are collaborating with some of the world’s top cardiovascular researchers to structure, analyse and publish the data, while incorporating the algorithms into a SaaS based product for clinical uses. You will be required to support and participate in Machine Learning trials, organising and analysing large volumes of echocardiograms. There will be opportunities to gain experience with a variety of Deep Learning classification and segmentation techniques in collaboration with some of the world's premier cardiology centres.

Project 1:

Information Extraction In the Legal Domain

At INTELLLEX, we handle millions of legal documents and we want to build an automated end-to-end system to extract useful information from the legal documents. For each document, we need to identify the areas of law being discussed and legal keywords being used.

In this project, you will be involved in the development of an automated end-to-end system - from data exploration, pre-processing, training and tuning Machine Learning classifiers, and running Machine Learning experiments to evaluate the performance of the system against standard benchmarks.

Project 1:

Forecasting of Financial Data Using Machine Learning

In this project, you will build a fast and reliable time series forecasting pipeline for sales data. This project will allow you to work on a big financial dataset, and evaluate internal and external factors affecting sales. This project will require you to assess various forecasting engines to identify the right model based on the data for the customer. The main aim of the product will be to provide our customer with the most accurate forecast with actionable insights.

Project 2:

Information Extraction Using Natural Language Processing

In this project, you will be developing an automated system which requires zero configuration or upfront annotations, using the latest developments in the fields of Computer Vision and Natural Language Processing for information extraction. You will work on the large dataset collected from the customer. The primary objective of this project is to automate the task of information extraction from data and process it in such a way that it can be utilised in our product.

Project 1:

Semantic Generative Compositions in Advertising

Pencil uses generative and reinforcement techniques to give machines a practical and usable understanding of what creativity is in the context of advertising - potentially the world's largest creative industry. In this project, you will be involved in specific challenges related to shipping a viable system for semantic generative compositions in the advertising space. Your goal will be delivering a working prototype with test users by the end of the project.

Project 1:

Predictive Analytics In Maritime

90% of the world trade travels via maritime, at least once in its lifetime. Portcast wants to apply Machine Learning to the extensive datasets from the maritime industry that it captures, and convert it into actionable insights. The project scope is to transform the raw datasets (from econimic indices, geo-spatial data, to news/sentiment information) into insights and patterns that would be interesting not just for the logistics industry, but to the finance and insurance companies as well.

Project 1:

Analysis of Human Biopsy Samples with Computer Vision

Cancer is one of the leading causes of death in the world and affects patients and their families for extended periods of time. Although there has been a significant amount of progress in treatment options, early diagnosis is still the main factor in determining the patient's prognosis. Qritive's mission is to enable doctors to diagnose cancers faster and more accurately, and consequently improving the patient's chances in fighting these diseases. To achieve this goal, Qritive is leading several joint projects to build these systems. The project collaborators are the two most prominent hospitals in Singapore, with a panel of expert clinical researchers and clinicians involved in these projects. The project is based on recent advances in Deep Learning and Computer Vision technologies. The goal is to reduce the time spent by doctors on the analysis of cancer tissue images and reduce the number of misdiagnoses.

Project 1:

Image Pre-Processing Through Deep Learning

Image pre-processing is a vital step to create segmented data of images for downstream processing. You will be involved in creating a state-of-the-art image processing framework by using Deep Learning methodologies such as Convolutional Neural Networks (CNNs) and bi-directional long/short-term memory units.

Project 2:

Contextual Data Extraction

Identifying and extracting the original context of the document is a significant step in automatic information extractions tasks. In this project, you will be extracting relationships between texts which are contextually related. This makes it easy to import the extracted data into a database or to provide it as a variable into an application.