• Building a multi-agent LLM platform using LangGraph and LLM API providers to orchestrate automated project onboarding,

scoping, and scheduling workflows

• Implementing controlled, deterministic agents with safety, legal, and scope enforcement and fallback model handling

• Automating orchestration and approvals using n8n workflows integrated with business systems and APIs

• Tagged and evaluated 18 controlled lab studies with 36 participants, employing mixed-methods analysis of surveys, transcripts, and interviews to uncover gender-based collaboration patterns.

• Demonstrated significant boosts in women’s self-efficacy and reductions in programming anxiety, especially in mixed-gender teams, highlighting the tool’s effectiveness.

• Identified changes in socio-emotional and behavioral dynamics, including increased trust, more democratic leadership styles, and improved collaboration equity across gender pairings.

• Produced empirical insights advancing research at the intersection of HCI, collaborative software engineering, and gender-inclusive tool design.

• Evaluated PairEquity, an inclusive real-time feedback tool using multimodal analytics to promote equitable remote pair-programming collaboration.

• Designed an end-to-end deep learning framework to predict synergistic effects of antibacterial drug combinations using pre-trained molecular graph transformers for drug representation and genome embeddings for bacterial profiling

• Integrated SMILES-based chemical descriptors, K-mer genomic features, and essential gene embeddings from Evo DNA foundation model to enable precise prediction of drug-bacteria interactions

• Developed different models such as a Bi-LSTM model with multi-head attention and data augmentation strategy, while working on the dataset efficiency and handling

• Engineered a smart wand system using NodeMCU, stepper motors, and a gear-driven mechanism, dynamically adjusting blinds based on real-time solar azimuth calculations from user-provided coordinates via Blynk mobile integration

• Developed a control algorithm that processes real-time data from a mobile device, computing solar azimuth angle achieving ±1° precision, and required blind rotation, ensuring accurate and automated adjustments

• Designed, developed, and integrated a robust Machine Learning pipeline for a real-time truck Automatic Number Plate Recognition (ANPR) system, leveraging tools and frameworks such as LabelImg, YOLO, CNN, Tesseract OCR, TensorFlow, PyTorch, OpenCV, and Python, tailored specifically for Weighmast weighing solutions. Achieved a high detection rate of 85%.

• Spearheaded the end-to-end development of the ANPR system, from model training to deployment, ensuring seamless integration with the existing Weighmast software for automated vehicle tracking and monitoring.

Seamlessly integrated the ANPR system with the existing Weighmast Software, streamlining operations and facilitating automated tracking and monitoring of vehicles.

• Developed and deployed a RESTful CRUD API using MySQL and Postman, enabling efficient, real-time database interactions, optimizing data retrieval and storage, and significantly improving operational efficiency and website functionality.