Daniel Firebanks-Quevedo Email & Phone Number

San Francisco, California, United States

Los Angeles, CA, US

Working on LLM research at Dr. Robin Jia's AI, Language, Learning, Generalization, and Robustness (Allegro) Labhttps://robinjia.github.io/

San Francisco, California, US

Armonk, New York, NY, US

- Trained and evaluated statistical, neural network, and transformer-based NLP models (e.g. for unsupervised text classification, sentiment analysis and entity recognition) that powered several IBM products such as Watson Discovery, resulting in $20m+ in revenue.- Led the design and implementation of critical features for the release of the embeddable.

Armonk, New York, NY, US

- Created a fault-injection framework for cloud-native applications to automate discovery of unobserved faults in staging environments, used by 10+ teams across IBM Research.- Built REST APIs for a conversational AI platform using Java and Elasticsearch, and increased integration testing coverage by 30%.

Chicago, Illinois, US

• Working with the World Resources Institute to construct a machine learning model that detects financial incentives for landscape restoration in policy documents, potentially reducing document processing time from.
• Developing an NLP framework in Python that offers summarization, entity recognition, topic modeling, and text categorization for policy analysts to extract relevant information and improve decision-making.
• Presented model and preliminary results to government officials from 5 Latin American countries, and submitted work to an international public policy conference.

• Helped humanitarian NGOs model the spread of COVID-19 in refugee camps or refugee camp like low resource settings.
• Developed a network-based, stochastic Extended SEIRS model.
• Integrated the model with a Flask app to be used by humanitarian NGOs

Oberlin, OH, US

• Designed and implemented an open, multi-agent cyber-security simulation environment for testing different scalable reasoning algorithms.
• Researched Partially Observable Markov Decision Processes with Dr. Adam Eck
• Created a framework to build intelligent bots through the Python (PySC2) andJava SC2 APIs that uses the Raw Interface of the game.

Oberlin, OH, US

• Engaged in weekly lab tutoring sessions for groups of 20-25 students, by assisting them in developing software design proficiency, identifying/resolving bugs, and peer programming.
• Classes: Introduction to CS, Data Structures and Systems Programming.

Oberlin, OH, US

• Provided first-level client services/support to students, faculty and staff by diagnosing software issues and formulating solutions to them.

Oberlin, OH, US

• Analyzed tweets from a large group of scientists for changes in sentiment and its possible effects on virality.
• Used Python and R to implement machine learning (Naïve Bayes, SVM) and natural language processing techniques.

Armonk, New York, NY, US

• Implemented a paraphrase generation framework in TensorFlow using deepreinforcement learning and submodular optimization for the training ofconversational AI systems.
• Adapted Transfer Learning and Seq2Seq abstractive summarization techniquesfor data augmentation of IT service requests.
• Surveyed and tested the performance of multiple text generation/decodingmethods on domain adaptation tasks with small datasets.

Washington, DC, US

• Trained a Conditional Random Fields (CRF) machine learning model to identify structure and extract text from accounting statements.
• Built tools that automate cleaning, merging, and query-based retrieval of financial/accounting data for text analysis and risk prediction.
• Wrote a parser in C# to extract text while maintaining logical structure from non-standardized HTML documents, increasing accuracy from 47% to 78%.

Master Of Science - Ms, Computer Science

Bachelor Of Arts - Ba, Computer Science, Minor In Mathematics

International Baccalaureate

Computer Science