About Me
I will graduate soon and currently surfing the job market for full-time positions.
Hello! I’m Sahil (pronounced as ‘saa-hill’), a grad student advised by Dr. Martin Swany in the Department of Intelligent Systems Engineering (ISE) at the Luddy School of Informatics, Computing and Engineering of Indiana University Bloomington (IUB).
A researcher working at the intersection of deep learning, distributed systems and systems for ML/ML for systems. The focus of my Ph.D. studies is on building efficient computation and communication models to scale artificial neural networks across edge, cloud and high-performance computing (HPC). Currently working on topics like distributed training, federated learning, model compression, communication protocols optimized for deep learning, stream processing and differential privacy.
Outside of research, I’m an amateur photographer mostly shooting wildlife and space. Here are some pictures.
Education
- Ph.D., Intelligent Systems Engineering, Indiana University Bloomington, USA.
- B.Tech, Electrical and Electronics Engineering, Guru Gobind Singh Indraprastha University, New Delhi, India.
Research Interests
- Deep Learning
- Federated Learning
- Distributed Systems
- Edge, Cloud and High-Performance Computing (HPC)
- ML for Systems/Systems for ML
- Intelligent Computing
Work Experience
- 2018-2024: Graduate researcher and Associate Instructor at the Luddy School of Computing, Indiana University Bloomington, USA.
- 2017-2018: Research Staff Member at DREAM:Lab in the dept. of Computational and Data Sciences, Indian Institute of Science, Bengaluru, India.
- 2016-2017: Data Scientist, RocQ Mobile App Analytics, HT Media Ltd., Gurugram, Haryana, India.
- 2015-2016: Big Data Engineer, Stayzilla, Bengaluru, Karnataka, India.
- 2014-2015: Software Engineer, Tatras Data Limited, New Delhi, India.
Publications
- Conference Proceedings
- Tyagi, S., & Swany, M. On Using Large-Batches in Federated Learning (in preparation).
- Tyagi, S., & Swany, M. Enabling Large-Batch Training via Learned Gradient Mapping (in preparation).
- Tyagi, S., & Swany, M. Flexible Communication for Optimal Distributed Learning over Unpredictable Networks. 2023 IEEE International Conference on Big Data (BigData ‘23), 925-935 (Accpt. rate 17.5%) [pdf].
- Tyagi, S., & Swany, M. Accelerating Distributed ML Training via Selective Synchronization. 2023 IEEE International Conference on Cluster Computing (CLUSTER ‘23), 1-12 (Accpt. rate 25%) [pdf].
- Tyagi, S., & Swany, M. GraVAC: Adaptive Compression for Communication-Efficient Distributed DL Training. 2023 IEEE 16th International Conference on Cloud Computing (CLOUD ‘23), 319-329 (Accpt. rate 20%) [pdf].
- Tyagi, S., & Sharma, P. Scavenger: A Cloud Service For Optimizing Cost and Performance of ML Training. 2023 IEEE/ACM 23rd International Symposium on Cluster, Cloud and Internet Computing (CCGrid ‘23), 403-413 (Accpt. rate 21%) [pdf].
- Tyagi, S., & Swany, M. ScaDLES: Scalable Deep Learning over Streaming data at the Edge. 2022 IEEE International Conference on Big Data (BigData ‘22), 2113-2122 (Accpt. rate 19.2%) [pdf].
- Tyagi, S., & Sharma, P. Taming Resource Heterogeneity In Distributed ML Training With Dynamic Batching. 2020 IEEE International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS ‘20), 188-194 (Accpt. rate 25%) [pdf].
- Widanage, C., Li, J., Tyagi, S., Teja, R., Peng, B., Kamburugamuve, S., Baum, D., Smith, D., Qiu, J., & Koskey, J. Anomaly Detection over Streaming Data: Indy500 Case Study. 2019 IEEE 12th International Conference on Cloud Computing (CLOUD ‘19), 9-16 (Accpt. rate 20%) [pdf].
- Chaturvedi, S., Tyagi, S., & Simmhan, Y. Collaborative Reuse of Streaming Dataflows in IoT Applications. 2017 IEEE 13th International Conference on e-Science (e-Science ‘17), 403-412 (Accpt. rate 36%) [pdf].
- Poster Presentations
- Tyagi, S. Scavenger: A Cloud Service for Optimizing Cost and Performance of DL Training. 2023 IEEE/ACM 23rd International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW ‘23), 349-350 [pdf].
- Tyagi, S., & Swany, M. Accelerating Distributed ML Training via Selective Synchronization (Poster Abstract). 2023 IEEE International Conference on Cluster Computing Workshops (CLUSTER Workshops ‘23), 56-57 [pdf].
- Journal articles
- Tyagi, S., & Sharma, P., OmniLearn: A Framework for Distributed Deep Learning over Heterogeneous Clusters, (under review) 2024.
- Chaturvedi, S., Tyagi, S., & Simmhan, Y.L. (2019). Cost-Effective Sharing of Streaming Dataflows for IoT Applications. IEEE Transactions on Cloud Computing, 9, 1391-1407 [pdf].
Skills
- Programming: Python, C, C++, Shell scripting, MPI, OpenMP, CUDA, Java, Scala, R, MATLAB, SQL
- Databases: MySQL, MSSQL, SQLite, HBase, Neo4j
- Frameworks/Libraries: PyTorch, TensorFlow, Keras, MXNet, Hadoop, Spark, Storm, Kafka, Elasticsearch, Slurm, Cloud APIs (GCP, AWS, Azure)
- Graduate coursework: High-performance big-data systems, Deep learning systems, Operating systems, Logic design in computer engineering, Introduction to Intelligent systems, Bayesian machine learning, Computer vision, High-performance graph analytics
Teaching Experience
Course |
Role |
Term |
High-Performance Computing (ENGR-E317/517) |
Associate Instructor |
Spring 2024 |
Computer Networks (ENGR-E318/518, CSCI-P438/538) |
Associate Instructor |
Fall 2023, Fall 2022 |
Operating Systems (ENGR-E319/519, CSCI-P436/536) |
Associate Instructor |
Spring 2023 |
Distributed Systems (ENGR-E510, CSCI-B534) |
Associate Instructor |
Spring 2022, Spring 2021 |
Cloud Computing (ENGR-E516) |
Associate Instructor |
Fall 2021, Fall 2020, Fall 2019 |
Professional Services
Awards
- NSF Student Grant: To present research at IEEE CLUSTER 2023, Santa Fe, New Mexico.
- Luddy Dean’s Graduate Student Award: In Fall 2023 for outstanding research.
- NSF Travel Award: To present research at IEEE/ACM CCGrid 2023, Bengaluru, India.
- Best early-career researcher poster award: Awarded at IEEE/ACM CCGrid 2023.
- Google Cloud Student Researcher (2021, 2022): Received GCP credits for research.
- Student Research Award: Funded via grant NSF Data Infrastructure Building Blocks (DiBBS) 17-500, for academic year 2018-2019.
Presentations and Talks
- 04/24: Guest lectures, “Parallel Computing with GPUs for Distributed ML Applications”, High-Performance Computing (HPC) course, Indiana University Bloomington, USA [pdf1], [pdf2].
- 12/23: Paper presentation, “Flexible Communication for Optimal Distributed Learning over Unpredictable Networks.” 2023 IEEE International Conference on Big Data, Sorrento, Italy [pdf].
- 11/23: Paper presentation, “Accelerating DistributedMLTraining via Selective Synchronization.” 2023 IEEE International Conference on Cluster Computing, Santa Fe, New Mexico, USA [pdf].
- 11/23: Poster presentation, “Accelerating Distributed ML Training via Selective Synchronization.” 2023 IEEE International Conference on Cluster Computing, Santa Fe, New Mexico, USA [pdf].
- 09/23: Invited talk, “Towards building efficient computation and communication models for distributed deep learning systems.” Mathematics and Computer Science (MCS) division, Argonne National Laboratory, Illinois, USA.
- 07/23: Paper presentation, “GraVAC: Adaptive Compression for Communication-Efficient Distributed DL Training.” 2023 IEEE International Conference on Cloud Computing, Chicago, Illinois [pdf].
- 05/23: Paper presentation, “Scavenger: A Cloud Service for Optimizing Cost and Performance of ML Training.” 2023 IEEE/ACMInternational Symposium on Cluster, Cloud and Internet Computing, Bengaluru, India [pdf].
- 05/23: Poster presentation, “Scavenger: A Cloud Service for Optimizing Cost and Performance of ML Training.” 2023 IEEE/ACMInternational Symposium on Cluster, Cloud and Internet Computing, Bengaluru, India [pdf].
- 12/22: Paper presentation, “ScaDLES: Scalable Deep Learning over Streaming Data at the Edge.” 2022 IEEE International Conference on Big Data, Osaka, Japan [pdf].
- 07/20: Paper presentation, “Taming Resource Heterogeneity in Distributed ML Training with Dynamic Batching.” 2020 IEEE International Conference on Autonomic Computing and Self-Organizing Systems, virtual [pdf].
- 11/18: “Real-Time Anomaly Detection from Edge to HPC-Cloud”, Intel Speakerships at SC18 (Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis 2018), Dallas, Texas, USA [pdf].
References
Available upon request..