Hi, I'mHemanth Kandula

Engaged in cutting-edge research and development projects funded by DARPA and IARPA, primarily focusing on advancing the state of Machine Learning (ML), Natural Language Processing (NLP), and Computer Vision (CV) technologies.

• • Currently working closely with a multidisciplinary team of researchers, engineers, and domain experts on IARPA HIATUS tackling challenging Authorship Attribution tasks, leveraging advanced NLP and ML techniques.
• • Currently developing advanced semantic segmentation models leveraging deep learning and global pose estimation for accurate 3D site modeling from satellite, aerial, and ground imagery in the IARPA WRIVA project
• • Played a pivotal role in the IARPA BETTER program and led the task of enhancing IE system by integrating sophisticated techniques for granular information extraction, question-answering, and textual entailment slot-filling, significantly improving BBN’s system performance.
• • Engineered a Transformers-based Information Retrieval (IR) re-ranker system for the IARPA BETTER program, which markedly improved the efficacy of BBN’s probabilistic IR system, leading to more accurate document retrieval.
• • Led a specialized effort under the DARPA HUGO program, overseeing the collection, curation, and assessment of Hindi datasets. Effectively communicated the accomplishments and progress to DARPA, ensuring transparency and alignment with project goals.
• • Maintaining an active role in contributing to BBN NLP repositories, helping in the refinement and evolution of the software stack

Managed projects that solved a few of the major unmet needs in human fertility, viral diagnostics with a diverse team of clinicians and engineers resulted in 14+ journal and conference articles.

• • Designed two state-of-the-art adversarial unsupervised and self-supervised domain adaption methods and investigated domain-shifted medical datasets. (Accepted for publication at Nature BME Journal.)
• • Created and curated various benchmark medical datasets of 500,000+ images for domain adaption.
• • Developed an automated Machine Learning framework for In Vitro fertilization (IVF) that outperformed embryologists.
• • Applied image processing and machine learning techniques to smartphone-based low-cost point-of-care Ovulation prediction device and deployed into an android application which predicts Ovulation at-home with 99% accuracy.
• • Proposed generalization method using unsupervised adversarial learning for smartphone-based COVID-19 diagnosis.
• • Developed web/mobile apps (Vue.js, Flask, Android) for medical image datasets acquisition and annotation, and for deploying ML algorithms that were used by clinicians from 9+ hospitals and health clinics in the US.
• • Built point‑of‑care low-cost diagnostics (<$1) devices interfaces with embedded internet of things (IoT) systems.
• • Lead & managed a team of research interns in applying computer vision and deep learning in medical imaging projects.

Worked at Electric Vehicle Engineering and Robotics (EVER) Lab in projects mobile and aerial robots

• • Deployed algorithm with semantic segmentation (SegNet) for indoor autonomous navigation in ROS instead of high‑cost Lidar sensors on CoroBot (mobile robot) which decreased costs by 40%
• • Designed and implemented end‑to‑end autonomous control for drone GPS navigation system for agricultural crop spraying

Primarily involved in designing and building Android applications. These apps were custom-made for university cultural festivals, aiming to enhance the experience for both participants and organizers. They successfully served a user base of over 10,000 students from various colleges across India, with Sastra University being our primary client for these projects.