Pooja Ahuja
Frontend DeveloperAI/LLMs
ServiceNow
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
AI/LLMs
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Crafting brain for machines https://github.com/pcomputo/
University of Nebraska-Lincoln
Master’s Degree with a Thesis, Machine Learning
N/A – Present
ServiceNow
Machine Learning Engineer
January 1, 2024 – Present
San Francisco Bay Area
Glean
Machine Learning Engineer
January 1, 2022 – Present
Palo Alto, California, United States
Aruba, a Hewlett Packard Enterprise company
Staff Machine Learning Engineer
January 1, 2019 – January 1, 2022
California, United States
Chewy
Senior Machine Learning Engineer
January 1, 2019 – January 1, 2019
JASK (now Sumo Logic)
Founding Engineer
January 1, 2017 – January 1, 2019
United States
University of Nebraska–Lincoln
Research Assistant
January 1, 2016 – July 1, 2017
Indian Institute of Technology, Bombay
Research Fellow and Teaching Assistant
July 1, 2014 – July 1, 2015
Mumbai Metropolitan Region
Barclays
Software Development Intern
May 1, 2013 – August 1, 2013
Automated time diary survey
February 1, 2016 – Present
Researching into replacing the human interviewer in time diary surveys.
Budgeted Learning in Bayesian Networks
January 1, 2016 – Present
The labels of the training data are given, but the features are each available at a cost, subject to an overall budget. We introduce new budgeted learning algorithms that exploit the improved representational power of Bayesian networks over naıve Bayes classifiers to improve purchase decisions. Finally, we explore refinements of these algorithms that exploit the concepts of d-separation and Markov blankets to further improve decisions and save time.
Modeling human learning in Multiagent system
January 1, 2016 – Present
Modeling human learning by doing and by observation, and incorporating these into the agent’s reasoning about how to acquire tasks for user-based collaborative human task assignment. Our solution develops an approach for modeling and learning uncertainty caused by environment openness to guide its decision making in maximizing human user reward and learning gains over sequences of tasks.
Rural IVRS
December 1, 2014 – March 1, 2015
An Interactive Voice Response System for the rural communities of India where farmers can share knowledge, communicate and get involved with their customers in trade without a middleman. Includes a Spring-based web application as well as Android apps for farmers and customers for easy and mobile access to the system.
Mobile app Classification and Recommendation engine
December 1, 2014 – Present
Worked on extreme multi-label classification for apps with Wikipedia as the label space, where training was done through active learning. An associative Markov Network of candidate categories is built, where per-category SVM classifiers act as one of the node features. The generated categories aided an app search procedure as well as personalized recommendations.
Video Categorization using a Localized Category Catalog
October 1, 2014 – July 1, 2015
The project involves assigning facets to videos in a video collection of a particular domain. A domain-specific label space, which is a subset of Wikipedia, is evolved and is used to tag videos in batches. Active learning is used to evolve this system, where user feedback is solicited through a Game-with-a-Purpose (GWAP) for the most uncertain facets of the most uncertain videos.
Pidgin plugins
March 1, 2014 – Present
The Windows-command plugin is a plugin for the Pidgin client. It helps you carry out various operation with the conversation window like changing the background color, toggling to fullscreen mode, etc. The send-option plugin for Finch and libpurple allows the user to send 1. Group list 2. Buddy list 3. Online buddies to the buddy selected.
GNOME patches
February 1, 2014 – Present
Coded two patches. The first patch is used to get the total number of search strings found in the whole document with efficient scanning and display it accordingly. The second patch is used for printing the print preview message and printing message accordingly. It displays the name of the present document, the present page of total pages being printed and a percentage of the completed document for the point the preview and the printing.
Pidgin patches
January 1, 2014 – Present
Coded three patches for Pidgin. The first patch adds a new feature called unsent messages. It keeps a track of all the unsent messages (messages typed but not sent). If the user happens to close the chat window that has the unsent messages then a warning message is shown accordingly. For the second on hitting a given search in the conversation log the total number of searches found in the present conversation log is displayed. However, if no match for the given search is found then "No match" is displayed. On an empty search field the name of the log along with its date is then displayed. The third change highlights the quick find search results. Also on hitting a new search the previous highlights are unmarked accordingly.
Generate performance statistics for metals
July 1, 2013 – Present
Coded the metrics for various database environments which generated the performance statistics for various metals. The metrics adapted themselves as per the fluctuation in market situation to give a precise result.
SQL server migration
June 1, 2013 – Present
- Worked on migration project of SQL Server 2003 to 2008. - Coded a tool in C# to locate the given component within 10,000+ files among all 40 databases. - Re coded many databases to support the new features as well as to remove the unsupported features. - Ran the advisory tool over 40 databases to check the migration success. Deployed it successfully.
Speech recognition, transcription for Linux/GNU for easy use (on-going research project)
April 1, 2013 – Present
The purpose is to develop a system that implements a Desktop Voice Control System in the Ubuntu Desktop viz. operate the whole Operating System using voice. The methodology used for Hand's Free Computing (alias name for the project) is implemented in C and Python in conjunction with CMU Sphinx, which is an open source tool, created to convert speech to text. The language model is within the dictionary that is coded in C. The acoustic model is invoked by Python script. SOX library is used for recording and Flac for dynamic compression. Hidden Markov Model assigns probabilities to words in its vocabulary list based on its analysis of the audio segments. Viterbi decoding searches the state space for the most likely state sequence that accounts for the input speech.
Microsoft's I unlock Joy 2
September 1, 2012 – Present
Developed 4 apps for Windows phone. The same are available for download from the Windows phone website.
Unique ID Management (IBM The Great Mind Challenge 2011)
November 1, 2011 – March 1, 2012
The motive of this project is to provide a unique ID to every citizen of India while providing them a birth certificate. Citizens after attaining the age of 18 must register at an RTO office, Collector office or Thasildar office and get a password to access the site. An ID card will be provided to every user after registration. Later, they can change their password. Once they enter the site they can pay their electricity bill, telephone bill, book railway tickets, airline tickets and pay taxes.
Electronic Market for IBM’s The Great Mind Challenge 2012
July 1, 2011 – Present
A web application, where users can buy vegetables and fruits directly from farmers. One can locate the nearest farmer also check for his specialty and reviewer stats. XHTML, AJAX, Javascript, Java, Goggle maps API used. We advanced to the third level of the competition and won the first prize for our college.
Brusky's Hexagonal Chess
January 1, 2010 – Present
Implemented Brusky's Hexagonal Chess in PROLOG to appreciate inherent AI-ability of the language.
Cultural Fit Analysis
The candidate's career trajectory is strongly focused on Machine Learning Engineering and AI research, with significant contributions in that domain. The target role is 'Frontend Developer', which represents a significant pivot from their established expertise. While the candidate has a strong technical background, the lack of explicit frontend development experience in their professional roles and projects (except for a mention of XHTML, AJAX, Javascript in one older project) indicates a poor cultural fit for a dedicated frontend role. Their project history shows a strong inclination towards research and backend/ML systems. The 'Rural IVRS' project mentions a Spring-based web application and Android apps, but the candidate's specific contribution to the frontend aspects is not detailed. The 'Electronic Market' project mentions frontend technologies, but it's an older competition project.
Soft Skills & Operational Fit
The candidate's project descriptions highlight initiative (e.g., 'First female MLE', 'Founding Engineer', 'Pioneered novel statistical approaches') and problem-solving skills. The diverse range of projects, from database migration to speech recognition and educational tools, suggests adaptability and a broad interest in technical challenges. However, the resume lacks explicit details on collaboration, leadership styles, or conflict resolution, making a comprehensive assessment of soft skills challenging. The candidate's experience is heavily skewed towards Machine Learning and AI, which is not a direct fit for a Frontend Developer role.