Bangalore, India

Nitish Gupta (Open to relocation)

(+91) 6280907473
nitish.cse2020@gmail.com
Knowledgeable Applied Deep Learning Researcher with over 2 YOE in the field linkedin/nitish-gupta

EXPERIENCE EDUCATION
Indian Institute of Science,
Amazon, Bangalore — Alexa AGI: Applied Scientist Bangalore - MTech (CS)
Aug 2022 - Nov 2023
2020-2022 (GPA: 8.3/10)
● Developed, tested, and deployed On-Device Speech2Intent (S2I) ASR model for a leading
luxury automotive manufacturer Panjab University,
Chandigarh - BTech (CSE)
● Achieved 50% word error rate reduction and 27% improvement in intent recognition
error rate for the S2I model 2016 -2020 (GPA: 8.8/10)
● Engineered scalable data pipelines processing 4000+ hours of 3P data, enhancing the
generalizability of Alexa ASR models by 32%
● Recognised for improving the alignment and overall quality of data with Voice Activity SKILLS
Detection and normalization techniques Deep Learning:
● Devised a gating metric with a projected 15% reduction in model rollbacks that led to a
robust production environment ● Automatic Speech
● Elevated code quality by actively engaging in and initiating multiple critical code reviews Recognition (ASR)
● Natural Language
Mastercard AI Garage, Gurgaon — Data Science Intern
Processing (NLP)
June 2021 - July 2022
● Large Language Models
● Implemented SOTA methods for time series generation using GANs, Encoder-Decoder,
and RNN-based architectures ● Generative Models
● Published in ECML PKDD 2021, analyzing SOTA time series generation methods ● Computer Vision
● Established a baseline model for smart payment retry for subscription-based services ● Recurrent Neural Networks
(RNN)
● Attention Mechanism
PUBLICATIONS ● Reinforcement Learning
Adversarial Generation of Temporal Data: A Critique on Fidelity of ● Diffusion Models
Synthetic Data — ECML PKDD, 2021 Programming and Tools:
Machine Learning and Principles and Practice of Knowledge Discovery in Databases
● TensorFlow
Conducted an in-depth literature review, led the implementation of state-of-the-art
● PyTorch
methods, and contributed to the paper writing process, shaping the overall research outcome
● Git
● Hugging Face
PROJECTS ● AWS
● pyspark
Diffusion Models for Image Segmentation (Master’s Thesis) — 2022, IISc
Customized unconditional Diffusion Probabilistic models for image segmentation, achieving
competitive performance comparable to GANs and SOTA encoder-decoder architectures ACCOMPLISHMENTS
CV claims check-worthiness detection (Eightfold.AI) —June 2022, IISc All India Rank 13
Developed and implemented a rapid, end-to-end ML system within 24 hours for CV-based Graduate Aptitude Test in
fact-checking, creating an annotation tool and employing Positive Unlabeled learning with Engineering (GATE)-CS/IT
BERT to achieve 93% accuracy on a crowdsourced dataset Score: 950/1000
Indic Language Question Answering (Kaggle) — Sept 2021, IISc Hackathon Winner
Implemented context-based question-answering system in Hindi and Tamil, employing (Eightfold.AI HACKIISc)
transformer-based models from Hugging Face library and achieving competitive results
Reliance Foundation
Product Category Prediction, Amazon ML Challenge — July 2021, IISc Scholar (1 in 40 from all
Achieved Top 1% ranking by accurately categorizing products from over 10,000 classes India)
based on provided title, description, and bullet points. District Topper (ICSE, Class
Deep-Q Learning with prioritized experience replay — May 2021, IISc Xth)
Implemented PyTorch-based Deep Q-Learning with Prioritized Experience Replay (PER) to
train an Atari game Pong-playing agent using visible screen frames
Smart Parking Solution — May 2020, PU
Achieved precise vehicle counting and number plate recognition for entering cars by
employing the YOLO algorithm for efficient object detection in Python