Vasudha Gupta Silvi

Search, relevance (recommender system), and content understanding: + Increase relevance and user satisfaction of search results and ranking by diversifying candidate generation, feature augmentation, and model improvements + Reduce friction and increase user engagement with improvements to autocomplete by incorporating user query intent (query classification, expansion, intent prediction) + Incorporate, as relevant, cold-start content (content without historical engagement data)

+ Served as the Tech Lead for a generative AI system producing rich content (long-form text) to improve user experience, engagement (MAU), and SEO; led C4 architecture design across domains (ML, Data, Backend, Web) + Developed a POC model for a user propensity to purchase concert tickets based on listening history, artist affinity, and proximity to venues; reduced data aggregation by 33% while retaining over 90% of sales

+ Designed, built, and A/B tested a statistical heuristic to dynamically allocate marketing budgets across multiple 3rd-party ad platforms, resulting in a 5% lift in dollar efficiency for user acquisition + Refined and implemented a multi-bandit arm approach employing Thompson Sampling to adjust advertising slots among a variety of ad styles based on historical performance

+ Management of multiple work streams, pipeline development, and client deliverables using Kanban + NLP pipeline development and augmentation for question-answer extraction and risk assessment of diverse financial document datasets; incorporation of confidence modeling; data sources management; client-facing UI integration + Flask app development for auto-verifying invoices + Failure data simulation and modeling using various probability distributions and Bayesian statistics

+ Evaluated different scene graph generation models (Iterative Message Passing, Factorizable Net, Deep Relational Networks) and applied Iterative Message Passing for domain-specific content

+ Developed a web app for capstone project: Predicting Survival Outcomes for ICU Patients (survival-prediction.herokuapp.com) using descriptive text data and employing term frequency-inverse document frequency (TF-IDF) + Awarded a fellowship that has a 2% acceptance rate as 1 of 12 in New York City's winter cohort and covers topics including data wrangling, predictive model design, and machine learning

+ Researched a variety of convolutional neural network (CNN) architectures to assess best fit for classifying and detecting small objects present in satellite imagery + Trained, tuned, and compared performance of two CNN architectures (SSD and Mask R-CNN) for detecting buildings in satellite images (23G from SpaceNet) using TensorFlow and Caffe + Developed, prioritized, and delivered software in weekly sprints with Scrum agile framework

Added value on client projects by increasing efficiency, decreasing loss, and simplifying processes using analytical and data-driven approaches: + Produced a KPI tool for one of London’s metro systems that funneled 200+ daily operations and activities to support senior management with data-driven decision-making + Coached 3 supermarket managers overseeing 27 stores to sustain process improvements in queuing, refilling, and restocking that increased revenue 10% across the region

Be More is using mindfulness to reduce unconscious bias: + Co-developed and managed the product development of reducing unconscious bias with mindfulness workshops + Secured $250K+ in seed funding to develop an online prototype with videos and exercises + Accepted to Brooklyn Community Foundation's Brooklyn Accelerator Incubator + Granted the Points of Light Service Award for an innovative approach with large-scale impact; video: bit.ly/BeMorePOL + Supervised 3 undergraduate interns responsible for industry research and social media presence Learn more @ BeMoreAmerica.org. Services include in-person workshops, online courses, and speaking events.

+ Implemented a $1.5M customer service improvement project with automated touch-screen kiosks and revised signage for 200K+ daily passengers at the 42nd Street Port Authority Bus Terminal + Managed the review and stakeholders of the $2.5B terminal replacement RFP at LaGuardia Airport + Developed a visualization tool for the C-Suite to review 500+ capital projects valued at $27B

I assessed the network capability, advocacy, and efficacy of a national association for its constituency for the "Aging in Place" movement, which allows seniors or older Americans (above the age of 55) to remain in their choice of residence for as long as possible. Following my research, I provided recommendations for expansion and growth for the Village Movement’s leadership as well as strengthening current relationships within network.