Machine Learning Engineering Intern at Georgian, building production-grade agentic systems and multimodal AI pipelines. My work spans LLM orchestration with LangGraph and Temporal, vision-language models for cross-camera re-identification, and self-improving agents deployed on GCP — with a focus on reliability, observability, and real-world impact.
As a Technical Facilitator for the Agentic AI Bootcamp at the Vector Institute, I worked with ML teams from RBC Borealis and TD to rethink how their AI systems handle complex, multi-step workflows. The bootcamp focused on moving beyond single-model chatbots toward architectures where multiple specialized agents collaborate, reason, and act in concert — a shift that demands both technical depth and a careful eye for reliability.
As a Technical Facilitator for the Multimodal AI Bootcamp at the Vector Institute, I mentored ML teams from TD, CIBC, and RBC in modernizing their document understanding pipelines. The bootcamp challenged industry teams to move beyond traditional text-only approaches and adopt state-of-the-art multimodal workflows capable of handling the complexity of real banking documents — dense with charts, tables, and mixed layouts.
During my Master's studies at Queen's University, I was part of Ingenuity Labs, a research hub focused on cutting-edge innovations in robotics and artificial intelligence. I had the opportunity to collaborate with fellow researchers across various domains, engaging in dynamic, cross-disciplinary projects. As part of the Robotics Cluster, I attended weekly meetings, where I interacted with roboticists working on a range of exciting projects. This collaborative environment enhanced my research, exposed me to diverse perspectives, and strengthened my understanding of the intersection between robotics, AI, and AR.
Under the supervision of Prof. Matthew Pan, I conducted research for my Master's thesis, focusing on the development of an Augmented Reality (AR) application. The project aimed to visualize the final design of under-construction buildings at their intended locations in real-time. This work involved integrating AR technologies with 3D models of urban infrastructure to enhance public engagement during the construction process. By creating an immersive experience for stakeholders, my research contributed to streamlining the planning and approval phases of urban development projects.
During the summers of 2023 and 2024, I supervised two summer interns at MItHRIL Lab, Queen’s University, guiding them through research projects focused on VR/AR. I provided mentorship on research methodologies, technical skills, helping them develop a strong foundation in their respective fields. Through regular discussions and progress reviews, I ensured a productive and engaging learning experience, fostering their growth as researchers while working on my own research project.
Over a four-month period at Queen’s University, I assisted a professor in designing and developing a new third-year undergraduate course in the MREN department. The course featured a hands-on challenge where students built their own wheeled robots for a Jurassic Park-themed competition. Their robots were tasked with grabbing dinosaur figures and storing them in designated boxes. My role involved structuring course content, selecting appropriate materials for building the robots, and creating a demonstration wheeled robot using these materials. Through this experience, I gained valuable insights into curriculum development, project-based learning, and effective teaching methodologies.
As the Women in Engineering representative in the GECE Student Council, I advocated for gender diversity and inclusion within the engineering community. I organized events and initiatives aimed at supporting and connecting women in engineering disciplines, fostering a collaborative and empowering environment.
As a Machine Learning Researcher at Avir, I specialized in image dataset curation, leveraging web scraping and advanced data cleaning techniques to build diverse, high-quality datasets. I also developed a user-based open vocabulary image hashtag prediction model—a task that involves mapping an image to its accompanying hashtags, which remains an open challenge in computer vision. Recent research highlights the importance of user behavior in building accurate hashtag prediction models. To address this, we designed a model that extracts a user representation based on their image history, allowing for more personalized predictions.
During my undergraduate studies at the University of Tehran, I pursued a major in Electrical Engineering and earned a minor in Computer Engineering. I took multiple courses in Machine Learning, Neural Networks, and Deep Learning, which strengthened my foundation in AI and data-driven modeling. These courses enhanced my ability to tackle complex machine learning projects. Additionally, I served as a teaching assistant for several courses, including Machine Learning, Signal Processing, and Control Systems, which deepened my understanding and reinforced my communication skills. I also joined the Communication Laboratory, where I worked on a watermarking project under the supervision of Dr. Akhaee, gaining hands-on experience in signal processing and digital security.