Mid/Senior Deep Learning Data Scientist
La Haus, Colombia

La Haus

A Latin American family’s most valuable asset is its home. Geographic and financial mobility is stagnated due to low liquidity, the absence of robust platforms and marketplaces, as well as fragmented and unprofessional service providers. Our mission is to radically transform the residential industry, through technology, data and world-class service: provide asset and geographical mobility to Latin American homes. To fulfill this mission, we are building a technological platform that assists the home buyer through their journey, ranging from the initial search, to financing to the final purchase.

The role

We are looking for a data scientist / machine learning modeler to join the Data Centered Solutions (DCS) development team at La Haus. This person would be working very closely with the Supply amp; Integrations team in the development and customization of machine and deep learning models and algorithms. The main purpose of these models algorithms will be to stream-line and automate big portions of our housing inventory information curation processes. This information includes a mixture of structured and unstructured data, including natural lenguage descriptions and close to two million property photos.

Mission:

The DCS team mission is to provide our home buyers with an experience that they would perceive as intelligent. In other words, customers should get a sense that the data we collect about them and their search preferences are handled in an intelligent way to improve their home-buying experience. The same should be true for the supply side, that is the inventory of properties that we are trying to match against buyers. The role’s mission is to devise new scalable strategies and algorithms, and maintain and improve existing ones, to both measure and improve the inventories quality, and when improvement is not feasible allow us to filter or down-rank low quality content.

Tasks and Responsibilities:

• Develop algorithms to automatically assess the quality and reliability of structured and unstructured information associated with a property-listing, visible to clients as a property detail page.
• Refine existing deep learning algorithms to assess the visual appeal of a property’s pictures, or construct new ones from scratch.
• Develop algorithms to automatically improve a picture maybe by automatically adjusting contrast, hue, cropping it, or applying a resolution enhancement (super-resolution algo)
• Develop algorithms to remove water-marks from pictures with minimal loss.
• Design new algorithms to generate with captions (alt-texts) for each property picture and classify them according to type of room or space they show.
• Create new algorithms for auto generation of high quality property descriptions and assessment of existing ones written by humans.

• Contribute to the continuous improvement of the team’s technical knowledge.
• Be in constant and close communication with various product managers to gain a deep understanding of product and business needs and how they could be satisfied with data science and ML tools.
• Be proactive, suggest simple, effective solutions that are easy to implement and maintain.
• Be propositive: suggest improvements in the data capturing, representation and storing procedures.
• Know and use methodologies for the development of predictive models CRISP-DM and SEMMA, paying special attention to aspects related to business understanding and integration.
• Stay updated on the latest developments and trends in the fields of ML, DL, Deep RL, as well as the technologies and architectures used for their deployment.
• Be empathetic towards PMs and the commercial side of the company, but firm in pushing forward a technical, data driven approach.
• Be didactic in the divulgation of what ML and predictive analytics can and cannot do.
• Ask smart questions, assume risks and defend new ideas.

Requirements

• Technical:
  • Python programming: high
  • SQL knowledge: medium/high
  • Deep knowledge in the use of at least one Python ML library such as scikit-learn or Turicreate
  • At least 2 years of experience developing ML models
  • Practical experience with a deep learning framework Tensorflow, Keras, Pytorch, FastAI or MxNet.
  • Familiarity with Amazon Sagemaker and Mechanical Turk, or any other cloud based ML provider is desirable.
  • Some experience building simple web applications to capture training data from humans would be a big plus.

• Communication:
  • Assertive and effective communication, with a didactic disposition.
  • Skillful code commenting
  • Superb reading comprehension in English and Spanish
  • Excellent

Job Rewards and Benefits