Future of Generative AI: Going beyond Text to Image
This blog highlights the recent achievements in the generative AI space, text to image and beyond and discusses what the future might hold for generative AI

CS Graduate from IIT Indore. Building AI systems, training models, and building production-grade systems that scale.
From research papers to production systems, I transform complex ideas into scalable solutions.
Yatharth Gupta is a Computer Science graduate from IIT Indore and an AI Infrastructure Engineer. He is known for his research on Diffusion Models and LLMs, including the development of SSD-1B (Segmind Stable Diffusion), a distilled version of Stable Diffusion XL with 2M+ downloads on HuggingFace. He also created SegMoE, a Mixture-of-Experts architecture for diffusion models. His expertise spans Kubernetes, Terraform, AWS infrastructure, and building scalable ML pipelines.
From writing my first line of code at 15 to managing GPU fleets at scale.
Started learning C from YouTube at 15. The beginning of everything.
Started undergraduate degree in Computer Science at IIT Indore.
Published "Keep Up" on the Play Store; my first shipped product.
Published paper on RL-based closed loop control of PSFB topology at SAE conference.
Built a robotic arm that plays Tic-Tac-Toe using computer vision and inverse kinematics.
Won Silver in High Prep NLP problem statement at Inter IIT Tech Meet.
Became President of the ML club at IIT Indore. Built the team, organized workshops, grew the community.
Finished 13th globally in the IBM Quantum Computing Spring Challenge.
Started as an ML Research Intern at Segmind, working on diffusion model optimization.
Released Distill SD; shrunk Stable Diffusion 1.5 to half the size while improving quality.
Built and deployed an AI robot outside the mess to help new students with campus questions. HOD of CSE shared it on LinkedIn.
Released SSD-1B; distilled SDXL to half the size with better quality. Hit 2M+ downloads on HuggingFace.
Led IIT Indore at Inter IIT Tech Meet. Managed 90+ people across 15 problem statements. Secured 6th place overall with 4 podium finishes.

Created SegMoE; combines multiple SDXL/SD finetunes into mixture-of-experts without training.
Worked with the Shutterstock founder on Dropshot; AI-powered social media. Built the ML pipeline and infrastructure.
Joined Fal as ML Engineer. Deployed 30+ models and optimized inference pipelines.
Conducted ML workshop for 100+ students at Fluxus; Central India's biggest techno-cultural fest.
Hosted the Inter IIT Sports Meet at IIT Indore.
Won second place in Pitch It Off entrepreneurship competition.
Switched to infrastructure team. Now own Fal Compute, App Files feature, and manage the GPU fleet.
Featured in a podcast discussing my journey in AI and engineering.
Graduated with a CPI of 9.16 from IIT Indore.
Represented Fal at the PyTorch Conference in San Francisco.
A selection of projects showcasing AI, infrastructure, and full-stack capabilities.

Technical insights, tutorials, and thoughts on software engineering.
This blog highlights the recent achievements in the generative AI space, text to image and beyond and discusses what the future might hold for generative AI
A blog that covers my early experience and notes on Reinfocement Learning, it explains the basics of reinfocement learning,
Click anywhere on the board to share your thoughts, feedback, or just say hi!
Take care! Great site!
Yatharth is amazing