S12 Favorite - Overcoming Broken Right-Sizing Models to Automate Real-Time Cloud Cost Optimization with Sharad Kumar & Harshit Omar, Co-Founders of FluidCloud

Episode Summary

Executive Summary: The episode follows Fluid Cloud co-founders Sharad Kumar and Harshit Omar as they explain why cloud and infrastructure migration remains painfully fragmented, costly, and slow. After discovering AI could not reliably generate correct Terraform for end-to-end migrations, they built a mapping engine for true click-button, zero-human-intervention cloud movement. The conversation covers product evolution, scaling, team culture, market timing around VMware/HashiCorp disruption, and startup lessons on focus, distribution, and resilience.

Main Topics: Why Fluid Cloud Exists (Priority: 5/5): The founders describe the pain of migrating infrastructure across clouds and accounts, especially after acquisition events and vendor lock-in. Their experience showed that consulting-heavy, manual migrations are too slow and expensive. AI Experiment That Failed (Priority: 5/5): They spent four months trying to use AI and custom models to generate accurate Terraform, but it could not produce 100% reliable migrations without human intervention. This forced a pivot to building their own mapping engine. Product Roadmap and MVP Philosophy (Priority: 4/5): Instead of shipping a quick patchwork MVP, they invested in automation, testing, and performance from the start so the product would be production-ready and scalable when launched. Scaling Architecture and Market Timing (Priority: 4/5): The founders argue they designed for Fortune 2000-scale multi-cloud from day one. They also see VMware/Broadcom and HashiCorp changes as major market catalysts that validate their timing. Team Building and Culture (Priority: 4/5): They emphasize hiring problem solvers, fearless builders, and people who believe in the mission. Their culture prioritizes discipline, collaboration, and avoiding toxic behavior. Lessons for Founders (Priority: 5/5): They warn against stealth-mode assumptions, trend-chasing, and underestimating distribution. Their advice stresses reality checks, persistence, and building for the user’s actual problem. Vision, Values, and Inspiration (Priority: 3/5): They discuss role models like Ratan Tata, Steve Jobs, and Elon Musk, plus their own commitment to giving back through a foundation and building a company with strong values.

Key Arguments: Cloud migration and infrastructure change are still overly manual and expensive, even for experienced teams. AI alone was not able to generate fully accurate, production-ready Terraform for migrations, so a deterministic mapping engine was required. The best startup ideas come from founders who are the super users of the problem themselves, not from generic customer interviews alone. A real MVP should be production-grade, testable, and scalable, not a fragile prototype with shortcuts. Market disruption from Broadcom/VMware and HashiCorp created a strong tailwind for a multi-cloud migration platform. Hiring should prioritize problem-solving ability, fearlessness, and alignment with company values over raw years of experience. Distribution matters as much as product quality; founders should build go-to-market strategy in parallel with the product. Startups require resilience and long-term commitment; success is a time-based outcome, not an instant event.

Data Points: Time spent on initial AI-based MVP attempt: 4 months - They tried and failed to use AI to generate accurate Terraform before pivoting to a mapping engine. Time to integrate after acquisition in prior startup: 8-9 months - Their previous acquisition experience revealed how slow and painful cloud migration can be. Core team at launch: ~10 people - They started with engineers from their previous company joining on day one. Current team size: ~35 people - The company has grown steadily since inception. Engineering team size: 26 people - Current engineering headcount mentioned during the scaling discussion. Unit test coverage at start: 10% - They said they invested early in quality and refactoring rather than shipping a brittle MVP. Performance benchmark scale: 35 million records - Their MVP was benchmarked for no-downtime performance at large scale. Sprint cadence: 2 weeks - They described a biweekly release cycle regardless of team size. Possible AWS account configuration combinations: 3-4 billion - Used to illustrate the complexity of building a complete cloud mapping engine. Estimated combined cloud configuration combinations across AWS, Azure, and GCP: 9-10 billion - Their rationale for why the problem is extremely difficult to solve exhaustively. Reported VMware price increase: ~1000% - They cited this as a major market disruption and catalyst for migration demand. Consulting cost example: $5 million - They contrasted consulting-led migration costs with the value Fluid Cloud aims to provide. Foundation contribution: 2% of revenues - They said the company gives 2% of revenue to the Fluid Cloud Foundation.

Pivotal Quotes: "AI was never able to come up with an 100% accurate Terraform which can just run on clickbait and migrate." — Harshit Omar: Explaining why the team abandoned its AI-generated Terraform approach after four months. "You can build a great company not being an asshole." — Sharad Kumar: Describing the cultural standard they want across Fluid Cloud. "Great product with good distribution. If content is king, distribution is God." — Sharad Kumar: Advice to founders on pairing product development with go-to-market execution.

Implications: Fluid Cloud is positioning itself as infrastructure’s answer to multi-cloud lock-in, with the market now primed by VMware and HashiCorp disruption. For founders, the episode argues for building from firsthand pain, prioritizing reliability, and treating distribution as a core product function.