Audrey Woods, MIT CSAIL Alliances | May 18, 2026
Richa Gupta has seen firsthand how digital augmentation can transform someone’s life. When her grandfather, an architect, struggled to adapt to new digital drafting software, Gupta built him a program to translate his hand-drawn sketches into three-dimensional digital models. “The glow in his eyes helped me understand how many times the human design intent gets lost because of the technologies at hand," she says. “It became my mission to bring technology and human communication together. It’s not about automation; it’s always about human augmentation.”
Working in Professor Randall Davis’s group, Gupta has spent her time at MIT CSAIL thinking deeply about how AI design can meet people where they are and adapt to a person’s workflow, not the other way around.
FROM ARCHITECTURE TO AI SYSTEMS
Gupta grew up between two worlds, with an artist mother (daughter of the architect) and a telecom engineer father. She describes her childhood as having “one hand with pencils, the other with keyboards” and jokes that the pencils won, since she ultimately got her undergraduate degree in architecture. But architecture was never just about buildings. “The idea of architecture is that you try to understand what a human is saying, and you retrieve that information and turn it into a tangible solution,” like a building plan. It didn’t take her long after graduation to realize where the communication between humans and their tools were falling apart, ultimately eroding the design intent and purpose of the field.
Gupta soon transitioned into a design technologist role, prototyping tools and workflows for designers using emerging technologies. She also co-founded ArchiDAO, a decentralized autonomous organization which brought together five hundred other architects and creative technologists around the world to share new approaches and ideas. These experiences gave her insight into how impactful such work could be, but she wanted to do more. So, in 2023, she came to MIT.
STEERING AI WITH REAL WORLD DATA
Gupta's research, which spans both EECS and MIT’s School of Architecture and Planning, centers on what she calls human-AI collaboration, or designing systems where AI augments human thinking rather than replacing or distorting it. She argues that AI should adapt to the user rather than force the user to adjust to the limitations and restrictions of the technology. "Tools should work for me. I should not work for a tool.”
Her thesis addresses a specific version of this problem in architecture and design, where designers using AI image generators tend to get tugged along by what the model naturally produces, using vague prompts like “green” or “sustainable” when, for example, trying to create environmentally-friendly building plans. Her system, which she calls Insight Informed Generative AI (IIGenAI), presents the user with real-world information—like the specific CO2 emission values or cost of materials—alongside the generated images, giving designers context which then steers their prompts toward more specific and targeted language. With that shift in user vocabulary and no additional fine tuning or retraining, AI outputs enable decision making in the workflows of real-world architects and engineers.
A related project called “Beyond Prompts” aims to bring the fullness of human intent into the prompting process. Today’s prompts are mostly blunt instruments, unable to capture the complexity of what a designer might have in mind when generating an image. Say an architect wants to design a building inspired by the ocean or waves. Working with her colleague Rohit Sanatani in Professor Davis’s group, Associate Professor Takehiko Nagakura in the Architecture department, and at the MIT Media Lab, Gupta has been exploring how to build such analogical structures and semantic wordings into the interaction to steer a model toward a specific output which captures a richer version of what the user had in mind.
For Gupta, the human-in-the-loop is the spine of any tool she builds. She’s seen too many computer tools launched on the assumption that people will figure out its parameters or adjust to its restrictions, losing precious data and potential in the process. Gupta believes the technology which will truly make an impact is designed, from the beginning, with the human in mind. “Human agency is key.”
FUTURE WORK & NEXT STEPS
Gupta is planning to re-enter industry this summer with newfound direction and experience in agentic AI, multi-modal AI, custom harnesses, compound AI systems, and more. She says she’s transitioned “from an architect of buildings to an architect of technological solutions” and is excited to expand the reach of her ideas. Really, her work going forward is not so different from trying to capture a customer’s vision in architecture, but now with AI models rather than steel and concrete.
When it comes to automation, Gupta is skeptical. "A lot of people are trying to automate things left, right, and center with agents, and we are losing more and more control instead.” She believes that in the long term, prioritizing human-AI collaboration at every level—emotional, psychological, engineering, managing, observability, etc.—will be critical to make these systems adaptable and allow for a synergistic adoption which preserves human agency and vision.
Gupta believes the real opportunity isn't applying AI to existing workflows but rethinking workflows entirely. Many problems previously believed to be unsolvable weren’t really unsolvable, just burdened with clunky tools and cognitive overhead. By combining the strengths of AI and human reasoning, and taking advantage of the multi-modal opportunities rapidly proliferating in AI tools, Gupta thinks the approach to almost all work should be reexamined from the ground up beyond the current and limited standard of mouses, keyboards, and screens. A world where people can communicate naturally and effortlessly with their tools will change not just how work gets done, but what work people attempt.
"No tool should define a human mind," Gupta says. "We can rethink all the processes and methods and communication and everything."
Learn more about Richa Gupta on her MIT Architecture page or LinkedIn page.
