Mar 2025 – Jul 2025 · 4 mos · Remote

Oct 2024 – Feb 2025 · 4 mos · Remote

Nov 2023 – Jul 2025 · 1 yr 8 mos · Hyderabad, Telangana, India

• Auditing team in Security org

Aug 2023 – Dec 2023 · 4 mos · Hyderabad, Telangana, India

• Tasks Platform team in Office Product Group

Dec 2022 – Aug 2023 · 8 mos · Hyderabad, Telangana, India

• Tasks Platform team in Office Product Group

Jul 2021 – Dec 2022 · 1 yr 5 mos · Hyderabad, Telangana, India

• Engineering Systems 365 in Office Product Group

Jan 2021 – Jun 2021 · 5 mos · Hyderabad, Telangana, India

• Office Fundamentals Security and Compliance Team

Oct 2020 – Jan 2021 · 3 mos · Bengaluru, Karnataka, India

• Developed a system for monitoring daily employee attendance using Java, Angular and MySQL.
• Developed scalable APIs to fetch data related to different assets and their metrics and handled raw data.
• Designed an analytics workbench and dashboard to monitor different assets and their metrics e.g. distance, estimated fuel, hours run, etc. over time.

Jul 2020 – Oct 2020 · 3 mos · Beirut Governorate, Lebanon

• Built a pipeline to integrate Named Entity Recognition and Relationship Extraction from legal documents using BERT for language modelling and BiDirectional LSTM model.
• Developed a module to extract clothing type and color from street style outfits using multilabel classification.

May 2020 – Jul 2020 · 2 mos · Hyderabad, Telangana, India

• Joined the Security and Compliance Division of Office Fundamentals Team, Microsoft Hyderabad.
• Aggregated data from azure storage queue and ADO Git REST APIs to collect metadata about bypassed pull requests and pushes.
• Implemented checks and automated code reviews to detect anomalies in office code check-ins. Created a Power BI dashboard to display the results.
• Analysed the associated data using statistical techniques, used t-SNE and PCA for visualization and implemented outlier detection algorithms like unsupervised KNN, Isolation Forests, etc.

Mar 2020 – Jun 2020 · 3 mos · Remote

• Worked on using AI to understand the repercussions on the vulnerable population during pandemics and enabling data-driven policy-making in collaboration with Project Work Groups and AI For Peace, UN.
• Managed a team of 10 people to create a dataset of policy responses against the Coronavirus pandemic, combining economic, healthcare and containment policies.
• Conducted data-driven impact-analyses on how the policies have affected vulnerable populations in terms of healthcare access, wage loss, employment loss and domestic violence.
• Correlated policies with vulnerable groups, evaluated government response and the effectiveness of each policy.

Oct 2019 – Dec 2019 · 2 mos · Remote

• Worked on detecting forest fire outbreaks to avoid disastrous deforestation in collaboration with Sintecsys.
• Wrote script to collect 100K+ landscape images with forest fires. Used hashing to remove identical images.
• Used CNN, transfer learning using ResNet and VGGNet for classification to get an accuracy of 95%.
• Implemented semantic segmentation using DeepLab V3 and UNet to detect segments of smoke in the images and attained a mean IoU score of 80.

Dec 2019 – Jan 2020 · 1 mo · Greater Allahabad Area

• Worked on multimodal emotion recognition, modality feature extraction and modality fusion techniques at Interactive Technologies and Multimedia Research Lab, IIIT Allahabad under Prof. Anupam Agrawal.
• Developed feature extraction techniques for facial expressions, body gestures, voice and spoken sentences from videos.
• Developed neural network architectures like transfer learning using CNNs and RNNs for emotion classification.
• Worked on decision level and feature level fusion for combining the different modalities.
• Evaluated performance of model on multimodal emotion datasets like FABO, CREMA-D, GEMEP, Polish dataset, etc.

May 2019 – Jun 2019 · 1 mo · Noida, Uttar Pradesh, India

• Worked on Multimodal Emotion Recognition on Polish data under Dr. Sridhar Swaminathan.
• Extracted the features of body motion, facial expression and speech from videos of Polish actors.
• Used neural network models to classify the data into one of the seven emotions in each modality of data.
• Combined all the modalities using weighted average approach to get the final result.

Mar 2019 – Oct 2019 · 7 mos · Gurugram, Haryana, India

• Developed a data visualization website using Python, Django, Matplotlib, Seaborn and Plotly.
• Using this website, a variety of interactive graphs can be built from tabular data, which is the input query dataset, and filters can be applied to the data for better analysis.