From Rockets to AI: Nishkam Batta in Conversation With Didi on Repeatable Aerospace Manufacturing

Interview by Nishkam Batta, Editor-in-Chief, HonestAI Magazine

Nishkam Batta sits down with Didi, co-founder and former CEO of Hera Technologies, for a conversation about building an aerospace manufacturing company, creating repeatable quality under AS9100, protecting people on the shop floor, and determining where AI can and cannot help manufacturers.

When Nishkam Batta asked Didi to introduce herself, the answer quickly became the story of a manufacturing company built from the ground up.

Didi co-founded Hera Technologies with her husband in 2015 after spending roughly two decades working with thermal technologies in aerospace. Her experience included multilayer insulation, fabrics, and materials used to control extreme temperatures in satellite applications.

Hera began in approximately 1,200 square feet with Didi and her husband as its two founders. Early work included supporting SpaceX's landing legs. The company subsequently expanded its capabilities around thermal protection systems and insulation used within rocket structures.

Within several years, the business had grown dramatically. Didi described expanding first into multiple units totaling roughly 40,000 square feet before eventually operating across two locations with approximately 100,000 square feet combined.

But this conversation with Didi wasn't limited to growth.

It was about what happens behind that growth: how manufacturers create repeatable quality, how leaders build processes around people, and what changes when AI enters that environment.

Nishkam Batta: What Does Quality Really Mean in Aerospace Manufacturing?

When Nish turned the conversation toward quality, Didi challenged one of the simplest assumptions about manufacturing performance.

Hera Technologies operated under an AS9100 quality management system, and Didi's experience taught her that repeatability was central to maintaining conformity as designs changed.

A component might conform to Revision A, she explained, while the same process might fail to satisfy Revision B, C, or D.

Her response was to become intensely process- and metric-driven.

When two operators following the same process produced different results, Didi didn't immediately assume one employee was at fault.

She questioned the process.

Were the work instructions clear?

Were the tools positioned correctly?

Was the employee properly trained?

Did the physical setup make sense for the person actually performing the work?

By auditing those variables, Didi reached an important conclusion about aerospace manufacturing quality control:

For Nish, who is exploring practical AI applications for manufacturers, that naturally raised another question: Could AI make those processes better?

Nishkam Batta: Where Could AI Make Aerospace Manufacturing Safer?

Before getting into AI, Didi explained the physical realities behind Hera's specialized manufacturing.

The company worked with materials including phenolics and silica. Machining them required careful attention to particulates, ventilation, filtration, personal protective equipment, fumes, and combustible materials.

Protecting employees, Didi explained, it wasn't simply about avoiding regulatory problems.

It was about protecting the people who supported the company.

By the time she exited Hera, Didi said the company had approximately 35 approved special processes across rockets and missiles.

When Batta asked how AI might support those special processes and the people responsible for them, Didi identified safety analysis as one of its strongest potential applications.

AI could read safety data sheets, analyze raw-material components, and potentially recognize adverse interactions much faster than people manually reviewing large amounts of information.

Looking back, Didi recalled dealing with combustible materials and fires while learning how different materials behaved.

AI, she believes, might have helped identify some of those risks earlier. For manufacturers, that creates a tangible AI use case: Give AI the documentation. Let it analyze the information. Ask it to identify potential risks. Then have experienced people validate what it finds.

Nishkam Batta: Can AI Improve Work Instructions and Repeatability?

Nish pushed the conversation further.

If AI can analyze documentation, could it also help manufacturers develop the procedures and work instructions necessary for repeatable quality?

Didi believes it can.

She pointed to something as simple as AI-generated meeting minutes.

Traditional meeting notes inevitably reflect the interpretation of whoever writes them. AI participating as a third-party recorder can create another representation of what happened.

She sees a similar opportunity around statements of work, procedures, and manufacturing documentation.

AI can organize information.

It can improve formatting.

It can potentially identify gaps.

And it can help manufacturers make their procedures more consistent.

But Didi drew a clear boundary.

AI does not possess the real-time manufacturing instincts accumulated through years of experience.

A system may understand dimensions or specifications conceptually, while an experienced machinist or engineer can look at a proposed operation and recognize that it simply won't work at a certain speed or under actual shop-floor conditions.

AI can analyze the process. Humans still have to validate the reality.

Nishkam Batta: How Do Manufacturers Bridge the Gap Between AI and Experience?

Nish then posed a harder question.

If AI lacks years of practical manufacturing experience, how do companies bridge the gap between what an AI system theoretically knows and what experienced people know from actually performing the work?

Didi's answer was unexpected:

Empathy.

She recalled meeting someone who did not discover that he was colorblind until around age 40.

Imagine, she suggested, that the same person was an operator following a procedure heavily dependent on color. He would interpret those instructions differently from another operator.

An AI system might generate a technically logical procedure without understanding that human reality.

That's why Didi believes manufacturers must validate not only whether an SOP looks correct but whether the people receiving it can actually execute it.

For Didi, validation therefore remains a human responsibility regardless of how deeply AI becomes integrated into a manufacturing business.

Nishkam Batta: Should Manufacturers Move Quickly on AI?

Nish acknowledged another challenge: getting AI to work reliably in precise environments can itself create additional work.

Didi agreed that gaps remain.

Her advice to manufacturers is to avoid attempting an overnight transformation.

She compared AI implementation to teaching someone to cook. You don't teach every recipe in a day. Some knowledge requires practice, patience, and accumulated instinct.

During her time at Hera, Didi said she created more than 65 work instructions. She wanted those instructions to establish clear lines for employees to follow, reducing uncertainty about what should happen and why.

That clarity creates accountability.

Accountability supports repeatability.

Repeatability supports quality.

AI can make that system easier to develop and maintain, but Didi does not believe it is autonomous enough to eliminate the experienced person. "It's a helpful tool, but it's not so autonomous that it doesn't need you."

Nishkam Batta: If You Were Running Hera Today, Exactly Where Would You Use AI?

Toward the end of the conversation, Nish made the question concrete.

He asked Didi to imagine she was still running Hera and had received a new project with a new scope of work requiring the company to develop a manufacturing process

Where exactly would she use AI, and where wouldn't she?

Didi said she would begin by giving AI the overall scope of what the company was trying to accomplish.

She would ask it to analyze the process and identify risks the team might not have considered.

She would also use it for formatting and to help ensure documentation followed appropriate guidelines.

But the core process flow?

The design decisions?

Those would remain human.

Experience still matters too much.

AI might identify a risk an experienced leader overlooked, and it might dramatically improve the clarity of documentation. But Didi would not allow it to become the sole source controlling the process.

For Nish it was manufacturing audience, whereas, Didi's answer establishes a useful boundary:

Use AI to analyze. Use AI to organize. Use AI to question. But keep experienced people responsible for validation and execution.

Nishkam Batta: Can AI Help Manufacturers Learn From Previous Projects?

Nish had one final practical AI question.

When a manufacturer receives a new scope of work, could AI search previous projects, statements of work, procedures, and solutions to find relevant experience?

Didi's answer was yes.

Rather than beginning every project from scratch, manufacturers could use AI to retrieve the knowledge embedded in previous work and use it as a foundation for new requirements.

For small and midsized manufacturers, this could become an important application of AI: turning years of scattered institutional knowledge into something employees can retrieve and reuse.

But Didi attached an important warning.

The past should be a foundation, not a constraint.

If manufacturers simply reuse existing procedures without challenging them, they risk being left behind as technologies and requirements change.

Her principle is simple:

Nishkam Batta and Didi: Manufacturing Is Ultimately About People

Although Nish's conversation with Didi explored AI, aerospace, quality systems, and manufacturing processes, a consistent theme kept emerging: People.

Didi's approach to repeatability starts with understanding the operator.

Her approach to safety starts with protecting employees.

Her approach to AI requires experienced people to validate its output.

And her approach to leadership is based on explaining why people are being asked to do something.

At one point in their conversation, Didi recalled telling an employee, "I need to talk to you on Monday," without explaining why.

The employee spent the weekend believing something was wrong.

In reality, Didi simply wanted to give the employee a W-2.

The experience taught her something deceptively simple:

The principle applies to employees, manufacturing procedures, organizational change, and increasingly to AI implementation.

The Future of Aerospace Manufacturing Isn't AI Versus Humans

Nish began the interview series with a broader purpose: to uncover meaningful examples of how manufacturers, particularly small and midsized companies can think about AI beyond the hype.

His conversation with Didi offers a practical answer.

AI doesn't have to autonomously run a production line to create value.

Manufacturers also shouldn't dismiss AI simply because it cannot reproduce decades of shop-floor experience.

There is valuable territory between those extremes. AI can review documentation, surface risks, organize work instructions, retrieve knowledge from previous projects, improve communication, and provide another analytical perspective. But final responsibility remains human.

That mirrors what Didi learned while building Hera Technologies: quality doesn't emerge simply because employees are told to produce quality.

You build processes that make quality repeatable.

AI can become part of those processes.

And perhaps the manufacturers that gain the most from AI won't be those trying to replace human expertise, but those that use it to make that expertise easier to capture, communicate, repeat, validate, and improve.

About Didi

About the Interviewer