AI Platform for Business

IBM Research pursues an open and integrated “AI first” approach to enterprise operations. Our cross-disciplinary research teams rethink enterprise architecture and transform business processes by combining AI algorithms, distributed systems, human computer interaction, and software engineering. We apply AI methodologies to accelerate the development of AI solutions from data acquisition, through model generation, to application delivery and continuous update. These include automated machine learning, continuous learning, flexible and dynamic composition, semi-automated model generation, visual analytics, human-AI interaction, and novel quality metrics.

The successful adoption of AI relies on the trust of end users, which means AI systems must be safe and adhere to ethical norms. We focus on delivering tools and techniques to identify and mitigate risks and violations, to help understand trust-related implications for and challenges of AI models, and to assist developers and data scientists in easily building trusted, safe, and explainable AI systems.

To let data scientists focus on models and data, we are innovating to create an AI platform that handles computation speed, scale, hardware selection, and placement. Advances in deep learning as a service, novel programming languages and programming models for AI, and elastic resilient deep learning at scale are examples of AI-optimized programming models and runtimes.

Featured Publications

Detecting Poisoning Attacks on Machine Learning in IoT Environments

N. Baracaldo, B. Chen, H. Ludwig, A. Safavi and R. Zhang

Slow and Stale Gradients Can Win the Race: Error-Runtime Trade-offs in Distributed SGD

S. Dutta, G. Joshi, S. Ghosh, P. Dube and P. Nagpurkar

Towards Extracting Web API Specifications from Documentation

J. Yang , E. Wittern, A. Ying, J. Dolby and L. Tan

Detecting Egregious Conversations between Customers and Virtual Agents

T. Sandbank, M. Shmueli-Scheuer, J. Herzig, D. Konopnicki, J. Richards and D. Piorkowski

Incremental Training of Deep Convolutional Neural Networks

R. Istrate, A. C. I. Malossi, C. Bekas and D. S. Nikolopoulos

Serving Deep Learning Models in a Serverless Platform

V. Ishakian, V. Muthusamy and A. Slominski

Runway: Machine Learning Model Experiment Management Tool

J. Tsay, T. Mummert, N. Bobroff, A. Braz, P. Westerink and M. Hirzel

Scalable Multi-Framework Multi-Tenant Lifecycle Management of Deep Learning Training Jobs

S. Boag, P. Dube, B. Herta, W. Hummer, V. Ishakian, K. R. Jayaram, M. Kalantar, V. Muthusamy, P. Nagpurkar and F. Rosenberg

DLPaper2Code: Auto-generation of Code from Deep Learning Research Papers

A. Sethi, A. Sankaran, N. Panwar, S. Khare and S. Mani

LSTMVis: A Tool for Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks

H. Strobelt, S. Gehrmann, H. Pfister and A. M. Rush

Generating Chat Bots from Web API Specifications

M. Vaziri, L. Mandel, A. Shinnar, J. Siméon and M. Hirzel

IBM Deep Learning Service

B. Bhattacharjee, S. Boag, C. Doshi, P. Dube, B. Herta, V. Ishakian, K. R. Jayaram, R. Khalaf, A. Krishna, Y. B. Li, V. Muthusamy, R. Puri, Y. Ren, F. Rosenberg, S. R. Seelam, Y. Wang, J. Ming Zhang and L. Zhang

Optimized Pre-Processing for Discrimination Prevention

Flavio Calmon, D. Wei, K. Ramamurthy, B. Vinzamuri and K. R. Varshney