cv
General Information
Full Name | Himanshu Thakur |
Location | Pittsburgh, PA |
hthakur [AT] andrew [DOT] cmu [DOT] edu | |
Languages | English, Hindi, German |
Education
- 2023
Master of Computational Data Science
Carnegie Mellon University
- Select Courses
- Large Language Models
- Multimodal Machine Learning
- Deep Reinforcement Learning
- Introdution to Deep Learning
- Introduction to Machine Learning (PhD)
- Cloud Computing
- Foundations of Computational Data Science
- Academic Services
- Reviewer for EMNLP 2023
- Core Committee Member, LTI Computing Insfrastructure
- Select Courses
- 2020
Bachelor of Computer Science and Engineering
Vellore Institute of Technology, Vellore
- Select Courses
- Machine Learning
- Artificial Intelligence
- Image Processing
- Neural Networks and Fuzzy Logic
- Parallel and Distributed Computing
- Virualization
- Data Structures and Algorithms
- Operating Systems
- Databases
- Networks and Communications
- Calculus
- Linear Algebra
- Statics and Probability
- Discrete Mathematics and Graph Theory
- Theory of Computation and Compiler Design
- Extracurricular Experiences
- Researcher at ACM-VIT
- Core Committee Member, SEDS-VIT
- Founding Member at Team Qubits
- Achievements
- G. D. Naidu Young Scientist Award, 2020
- Sir. M. Visvesaraya Award, 2019
- Secured 1st position in 7 national hackathons
- Ranked in Top 3 teams at 3 national hackathons
- Secured 2 patents
- Select Courses
Research Experience
- May 2023 - Aug 2023
Research Scientist Intern
Abacus.AI
- Conceptualized and led research on exploring linear properties of LLM adapters (soft prompts, LoRAs)
- Invented a novel learning algorithm to approximate a linear combination of pre-trained LoRAs to enhance generalization and multi-task performance
- Collaborated with researchers and developed new metrics for debunking staleness in popular NLP benchmarks due to memorization in LLMs
- Jan 2023 - July 2023
Graduate Research Assistant
Robotics Institute, Carnegie Mellon University
- Led and mentored a team of 6 research assistants for increasing generalization and robustness of computer vision tasks (segmentation, classification and tracking) for robotic e-waste disassembly at Biorobotics lab
- Devised a multimodal deep neural network architecture for semantic segmentation
- Developed a new loss function robust to label noise, improve baseline performance by 15% IoU and generalization 20% IoU
- Implemented an out-of-distribution detection algorithm achieving 98.6% accuracy
- Feb 2021 - Feb 2022
Research Intern
SketchX Lab, University of Surrey
- Invented a novel active learning algorithm for fine-grained cross-modal instance-level retrieval tasks,
- Increased mean top1 accuracy by 5% over state-of-the-art technique, published a first-authored paper to a top-tier conference (BMVC).
Work Experience
- Aug 2020 - July 2022
Senior Technology Associate
Morgan Stanley
- Developed a distributed deep learning framework for real-time failure prediction (time-series forecasting) and detection;
- Led to 5000x reduction in failure detection time, enabled prediction of batch job failures with 94% accuracy, awarded best contributor to global firm resiliency out of 5000+ nominees.
- Pioneered graph database as a service to enable 100x faster search and data lineage applications
- Rsearched and developed a domain-agnostic recommendation engine (using large-scale and distributed training of deep learning models) powered by graph databases, recognized as most innovative firmwide project.
- Dec 2019 - July 2020
Data Scientist Intern
Locale.AI
- Developed a novel clustering algorithm for human activity discovery from raw geo-spatial ping data using semi-supervised representation learning, enabled automated discovery of 18% new business areas from previously unused data.
- Productionized and scaled a timeseries anomaly detection framework using temporal convolutional networks (TCN), increased requests throughput by 6x, helped 10 companies in reliably predicting vehicle vandalism and trip delays.
- Jan 2020 - June 2020
Machine Learning and Web Development Intern (Part-Time)
S4S Technologies
- Worked on creating the entire internal operations dashboard for the team as well as a computer vision solution to solve food quality assurance problem.
- May 2019 - July 2019
Summer Intern
Sigmaways Inc
- Worked on AI based portfolio recommendation engine using LSTM based predictor and Meta-Heuristic Optimiser. Also, built a chatbot interface to the service.
- May 2018 - July 2018
Software Engineering Intern
Nova (P & D)
- Modernized the existing software capabilities and introduce a new method for speedy billing.
- Worked on Data Analytics and Software Development to create a new Edge-Billing system which allowed multiple working staffs to finalise a bill on their smartphones itself.
- Allowed multiple simultaneous billing, hence speeding up the process and developed software features which allowed better insights into sale and purchase.
Open Source Projects
- 2020
AI on the Beach
- Developed a data visualization library that can render over 10 GB of data into a single animated scatter plot under minutes.
Honors and Awards
- 2020
- G. D. Naidu Young Scientist Award
- 2019
- Sir. M. Visvesaraya Award
Other Interests
- Hobbies: Music Composition, Hiking, Travelling