Click on a project's image for more details.
Domain: Commercial real
estate brokerages.
Problem: Client team
members in these brokerages need to share and organize files within
a web application that helps them schedule and structure a real
estate deal's progression in a consistent way. To do so, they
submitted files to a back-office data team, which verified data
integrity. However, they were unable to see the status of submitted
file processing or consistent reasons behind various file submission
rejections.
Solution: Led a team of
three software developers to construct an uploaded file processing
status view in the web application's email inbox for the real estate
deal teams, which linked up in real time to matching back-office
file processing capabilities enhanced by the same team of
developers. See the photo carousel on the left for a visual
demonstration of both the client-facing and back-office facing
portions of this project.
Domain: Sales campaigns
using e-mail and LinkedIn messages.
Problem: Clients could
manually rate a prospect's overall "sentiment" based on their
responses to a given series of pre-programmed emails or messages,
but this was somewhat time consuming and potentially tedious.
Solution: Two separate
solutions for automated sentiment analysis of customer text
responses were explored. The first approach used a pre-existing
commercial API for sentiment analysis, which could be fed a given
text and returned a sentiment. When this fell through due to a lack
of communication from the commercial API provider, a second approach
was explored. This latter approach used an open source latent
semantic analysis (LSA) natural language processing technique, which
was fed thousands of previous real customer responses and their
manually assigned sentiments to train the program for automated
sentiment assignment of future customer messages.
Domain: Sales campaigns
using e-mail and LinkedIn messages.
Problem: The software
made use of huge interactor objects, that is, one or more sequential
actions modeled in code (e.g., generate an email, send it out) that
could have dozens or hundreds of inputs. The quantity of this data
made it infeasible to log in a storable format, in its raw form.
This meant that the engineering team could not search production
logs for errors.
Solution: A custom
logging interface to provide an easy to use logging API for
developers. It allowed for data inputs of interest (e.g., a user's
ID) to be filtered and logged alongside all accompanying interactor
events involved with the data. The end result was searchable error
logs for what were previously opaque bugs found buried in a chain of
past events.
Domain: Hiring software
for small and medium businesses.
Problem: The software
was being used internally by employees at a recruiting company for
creating and editing job posts. Many of these posts are similar in
structure and content, especially for positions at larger companies
and repeat clients. Greater consistency and lesser tedium was
desired in the repetitive creation of job posts.
Solution: A database of
shared, curated, editable "snippets" of parts of job ads was
developed, along with an associated user interface. This sharable
"lego kit" approach allowed for standardized, rapid write ups for
job postings.
Domain: Hiring software
for small and medium businesses.
Problem: The software
integrated with numerous application feeds of different hiring sites
such as Indeed, LinkedIn, Monster, etc. Facebook was rolling out its
own job site, and an integration was desired.
Solution: Integration
with the then nascent Facebook Jobs API, requiring liaising with the
implementation team at Facebook in order to push new job postings
from our hiring software to their site, and the corresponding
webhooks to receive job applications from them.
Domain: Hiring software
for small and medium businesses.
Problem: The software
featured the ability for users to create and edit their own job
postings, prior to having the software place them onto numerous job
portals. One add-on portion of the job posting feature was a
psychometric section, which allowed job posting creators to choose
from sets of personality screening questions, or to even add their
own. However, this feature was quite outdated and clunky at the
time, requiring an update.
Solution: Substantial
enhancement to the platform’s job application workflow.