Sudhendra Kambhamettu

An AI Product Developer focused on building relevant AI systems.

Boston, Massachussets, USA

LinkedInX
SK

About

As an AI Product Developer and DL Engineer, I have a proven track record of pioneering in the field of artificial intelligence, especially in computational biology and deep learning. My expertise lies in leading and collaborating with diverse teams, fostering an environment that encourages innovation and peak performance. Currently, my work primarily involves PyTorch, Huggingface and cloud tools like AWS, GCP, and Azure. With over 4 years of experience, including international collaborations, I have demonstrated exceptional skills in developing and implementing cutting-edge AI solutions across various domains.

Work Experience

Camp4 Therapeutics
Hybrid

Jul. 2023 - Dec. 2023

Computational Biology Co-op

Led a deep learning team in enhancing RNA nucleotide reactivity prediction using Large Language Models (LLMs), achieving a notable 0.17 Average MAE for RNA reactivity Prediction. Developed a scalable AWS framework for genomics data, improving the AUPR benchmark by 50% and introducing visualization techniques to increase LLM explainability in RNA analysis. Pioneered proprietary tokenization algorithms for embedding coherence.

Northeastern University
Hybrid

Feb. 2023 - Jul. 2023

Research Assistant

Developed a Universal Multimodal Encoder with late fusion, hierarchical fusion and early fusion methods combining DINO pre-trained ViT model and BERT for image and text modalities respectively. Evaluated on Hateful Memes Dataset (0.85 ROC-AUC) & private twitter mental health dataset (95% accuracy).

RAPYD AI Gmbh
Remote

Dec. 2020 - Dec. 2021

AI Product Developer

Played a key role in a startup's foundational team, leading the development of unique Computer Vision and NLP applications. Architected the rapyd.ai API, integrating AWS, GCP, and Azure, to provide a streamlined, no-code AI solution.

Education

Northeastern University

2022 - 2024
MS in Artificial Intelligence.

Vellore Institute of Technology

2018 - 2022
B.Tech in Computer Science with a specialization in Data Science.

Publications

Cerebro

Feb. 2023
Kambhamettu, S., Vimal Cruz, M., Anitha, S., Sibi Chakkaravarthy, S. and Nandeesh Kumar, K., 2023. Brain–Computer Interface‑Assisted Automated Wheelchair Control Management–Cerebro: A BCI Application. Brain‑Computer Interface: Using Deep Learning Applications, pp.205‑229.

Skills

Python/Java/C++
SQL/BigQuery
Bash
Pytorch/Tensorflow/Keras/Huggingface
Numpy/Pandas/Scikit-learn
Spacy/Nltk/Open-cv
Matplotlib/Seaborn/Tableau
AWS/GCP
Docker/git

Projects

LITART

Automatic & Scene consistent Generation of Textually Accurate Fictional Characters from Novels.

Research Project
Python
Pytorch
NLP
Generative AI
Huggingface

Scrap Tire Forecasting & Supply Chain Optimization

Achieved award‑winning (2nd Place) performance for forecasting scrap tires and optimizing supplychain for reduced carbon emissions towards the Reinvented the wheel 2.0 hackathon.

Hackathon Winner
Python
Timeseries
NLP

GODS (Guiding Object Detection using Segmentation) Pre‑training

An innovative pretraining strategy for OWL-ViT v2, integrating segmentation and attention objectives to significantly enhance its generalization across unseen categories, achieving a notable increase in mAP to 46.6 for novel object detection on the LVIS dataset.

Side Project
Pytorch
Self-supervised learning
End-to-end framework

HumanizeXAI

An intuitive web-tool for analyzing & evaluating AI models for vulnerabilities and producing interpretable statistics.

Course Project
Python
Interpretable AI
streamlit.io

Harmful Algal Bloom Detection

Detects harmful algal blooms in lakes from satellite Images. This project was in collaboration with NASA.

Side Project
Google Earth Engine
Microsoft Planetary Computer
Jupyter Notebook
BigQuery

Monocular Image Depth Estimation using Vision Transformers

Improved depth estimation with ViT-tiny 16, achieving a low root mean squared error of 0.138 and high processing speeds up to 22 fps, outperforming current techniques and enabling efficient real-time 3D reconstruction on complex datasets like NYU-depthV2 and CITYSCAPES.

Course Project
Python
Computer vision