Machine Learning Engineer (Part-time) at Folobotics

Led the machine learning team to design a multimodal AI system for a humanoid robot using Agile methodology, combining speech recognition, LLM-based dialogue, and vision-based human activity recognition; mentored an intern in successfully building the visual AI interface employing ViTs. Developed a contextual conversational AI using Gemini and Hugging Face Transformer models, integrating the Langchain framework to orchestrate RAG pipelines, manage prompt templating, and facilitate tool-based interactions with external data sources for improved relevance and response grounding. Built text processing workflows involving tokenization, vector embeddings, and prompt engineering to enhance the model’s understanding and coherence in dialogue. Deployed a real-time web interface via Streamlit on Hugging Face Spaces, containerized with Docker for scalable and efficient delivery.

Graduate Research Assistant at Carleton University

Developed a cytology cervical cell image segmentation and separation pipeline using GAN, Transformers, and fine-tuned CNNs in TensorFlow & PyTorch. Achieved a 22% improvement in separation performance over prior benchmarks, with IoU = 0.97, Dice = 0.98, and 99% segmentation accuracy. Built dashboards in Python and Power BI for real-time monitoring, delivering insights to supervisors. Maintained code version control with Git. Published research in IEEE International Symposium on Medical Measurements and Applications 2024, funded by NSERC.

Graduate Teaching Assistant at Carleton University

Taught and coordinated labs for Data Management with Python, Computer Systems Architecture, Digital Communications, and Data Structures & Algorithms, mentoring students in Python, C, MATLAB. Enabled a 30% increase in student performance through individual doubt clearing sessions for assignments, detailed feedback, technical support, and structured evaluations.

Generative AI Associate at Innodata

Evaluated and benchmarked Meta AI LLM model outputs for linguistic accuracy and logical reasoning. Enhanced dataset quality through various analysis techniques, ensuring high-quality training data for model fine-tuning. Employed prompt engineering and conducted content quality assessments.

Graduate Research Assistant at Carleton University

Developed a cytology cervical cell image segmentation pipeline using GANs and Transformers. Achieved significant performance improvements and published research findings.