August 2, 2026, (Inside AI) — An MTech student at IIT Hyderabad has built a hybrid physics-ML screening platform that slashes cold-forging die design costs for small manufacturers, winning a gold medal and a job in Japan in the process.
Guna Shekhar Reddy, a final-year student in the Techno-Entrepreneurship programme, developed the tool after witnessing firsthand how 85% of India’s casting and cold-forging SMEs lose 10-15% to defect-driven cost overruns. His platform screens die designs before full simulation runs, catching likely failure modes early and compressing iteration cycles that typically take two to four days.
The work earned him the Academic and Co-curricular Excellence Award at IITH, a gold medal for balancing strong academics with cross-domain co-curricular work. It also landed him a position at a Japanese firm, where he will continue developing tools that bridge physical manufacturing and machine learning.
Reddy’s journey began not in a lab but on shop floors. He visited casting and cold-forging industries near the IITH campus in Sangareddy, approaching 10 companies directly. “When people ask how I got design engineers to actually talk to a student, the answer is simple: I went and asked,” he said. Fieldwork included onsite research at Vedantha Tools, Tyche Diecast, and Eqic Industries, with trips to Pune, Chennai, and Bengaluru.
Read: MIT JARVIS Challenge Reveals AI’s Limits in Jet Engine Design
The Cost Barrier That Breaks Small Shops
The core problem is economic. Simulation software that could prevent defects early costs upward of ₹1 crore annually, a figure out of reach for most SMEs. As a result, a shop capable of serving 40–50 clients a year ends up serving 20–30. Reddy’s platform sidesteps that barrier by sitting atop existing workflows, screening designs before committing to expensive full simulations.
“The hardest part right now is adoption — the platform sits on top of what teams already do, so the value is not obvious until someone has actually used it,” Reddy noted. The prototype, which won a ₹20,000 IITH_BUILD grant from a cohort of over 100 pitches, is now in industry validation with SME design teams testing it against real die geometries and defect data.
This approach mirrors broader trends in physics-informed machine learning, where hybrid models reduce computational costs by embedding known physical laws into neural networks.
From a Village to a Gold Medal
Reddy’s path to IIT Hyderabad was unconventional. Born in Nagaram village in Suryapet district, he grew up in Hyderabad and studied at Johnson Grammar School under the ICSE curriculum. His father, who worked in watch distribution for HMT and Sonata, instilled a philosophy that Reddy credits for his interdisciplinary success.
“My father had a line he used often: ‘Jack of all trades, master of none, still better than master of one’,” Reddy recalled. “He believed that a person who could move across domains, that is, someone who refused to be boxed in by a single skill, would always have more to offer than someone who only ever went deep in one direction.”
After a BTech in Mechanical Engineering at Vasavi College of Engineering with a 7.27 CGPA, Reddy cracked GATE during the pandemic while diving into entrepreneurship literature. His CGPA at IITH now stands at 9.51. “In undergrad, I simply did not understand what a CGPA means or what it shapes later,” he said. “At IITH, that clarity came early.”
The gold medal recognizes not just grades but co-curricular breadth. Reddy stayed involved in multiple clubs, picking up working knowledge across fields. “I found I could stay involved in a few clubs at a time, pick up working knowledge across different areas, and let each one inform the others, exactly the way my father had always described a life well-lived.”
Read: InfiMaker Launches AI-Powered Desktop 5-Axis CNC Machine
Reddy’s upcoming move to Japan marks a new chapter, but his focus remains on the shop floor. The platform’s success will depend on convincing SMEs that a lightweight screening tool can outperform intuition without disrupting daily operations. For an industry where margins are thin and trust is earned through results, that may be the hardest simulation of all.
