Q&A: Sriram Haran, Managing Director and CEO of KeepFLying

Q&A: Sriram Haran, Managing Director and CEO of KeepFLying

Sriram Haran is a first-generation entrepreneur, who started at the age of 21 with academic background in manufacturing engineering (Singapore) and computer science (U.K). After a foray into multiple industries such as mining, education, trading, manufacturing and technology, he ventured into aviation during the pandemic. He has built ventures across the globe from Australia to the Americas over two decades. As the managing director and CEO of KeepFlying, he handles go-to-market initiatives and the overall leadership of the company. Editor-in-chief of Aerospace Tech Review, Joy Finnegan, had the opportunity to speak with him recently to learn more about AI, KeepFlying and the company’s plans for the future.

ATR: Everyone seems to be talking about AI right now. How do you see AI changing our world?

Haran: The recent buzz around ChatGPT and other large language models convince me that anything that doesn’t kill you can only make you stronger. I think we as humans will embrace this to become more productive and make more meaningful contributions to our lines of business.

This is certainly a paradigm shift and we’ll look back on this in retrospect in the future and say that ethically, we made the best use of it. While I am aware of the dangers of these tools, as evidenced by those who’ve spoken about it, I am optimistic that most of this will be put to good use.

ATR: What about AI in aviation – how will AI impact our sector of business?

Haran: Aviation is one of those industries where, despite technological advancements when it comes to flying, a lot of the maintenance and records-keeping processes are still very manual and laborious.

Despite efforts by ATA, AWG and IATA to come up with strategies to digitize these processes, we’ve seen various levels of adoption from simple CMS tools all the way until Tier-1 predictive tools like Skywise, Aviatar and GE Predix. Making sense out of unstructured data to help make maintenance and commercial decisions is where AI tools can play a great role in easing the burden of labor-heavy functions.

jet in hangar

ATR: Why was this company, KeepFlying, formed?

Haran: We sensed that there was a lot of focus on predictive maintenance whereas in reality, during the course of the pandemic, very little was being done to build tools that used the underlying airworthiness and maintenance data of an aircraft, engine or component to build financial/commercial profiles.

Our question was, “How can one visualize the commercial impact of a decision being taken against an Aircraft or an Engine before taking it?”

In an era of Digital Twins, we ended up building the first Digital Financial Twin — we christened it the FinTwin.

ATR: Is KeepFlying an AI-based product?

Haran: Very much, when dealing with unstructured data, it is important that any data model we built was Asset-type specific. This means that the model recognizes the nuances of an A320, or a B737 or a CFM56-5B or a V2500 when creating commercial profiles of each asset. The fundamental principle to this is what the world is talking about today – large language models. We developed LLMs with aviation specific semantic layers when we launched the company.

ATR: There are two editions of your product: 1. MRO and 2. Asset Owner. Tell us about the differences.

Haran: The FinTwin MRO edition allows engine & airframe MROs to simulate hangar & shop visit profiles from slot sales to induction all the way until delivery/redelivery using a combination of market data and asset-specific models. In essence, by running multiple simulations from slot prospecting to proposal to contracting until the slot sale, data sets can gradually be added with every iteration to accurately predict manpower and material demand well before an Aircraft or engine enters the facility.

The FinTwin Asset Owner edition allows lessors, financiers & operators to simulate residual values, remaining useful life, redelivery risks and costs as a factor of projected utilization, environment and operational data.

This can help with simulating decisions around shop visit build goals, lease rentals and reserves, trade decisions (lease extension, return, scrap/part-out), module swaps and exchanges among others, and visualize commercial impact.

ATR: Each of those two editions is also broken down into separate sections. Define those for our readers.

Haran: It is further broken down as Aircraft and Engine editions. Depending on whether you are an aircraft or engine lessor, operator, financier, airframe or engine MRO, one or both of these editions might be relevant to you.

ATR: Your company info says: “KeepFlying ascertains financial and risk forecasts in a post pandemic era where creditworthiness and legal hassles can consequentially present pitfalls when repossessing an Aircraft or engine in case of an event of default.” Talk about how KeepFlying does this.

Haran: In addition to airworthiness and maintenance data that help ascertain residual values, help forecast costs and reserves, the underlying creditworthiness of an sirline and the jurisdictional risk that the country entails (a factor of whether the country has ratified the Cape Town Convention or not) allows the platform to weigh in factors that may affect the cashflow forecasts against assets or predict an event of default. Some of these risks and the costs associated with them are fundamental to any risk of repossession that may add burden to the lessor in case of an event of default.

ATR: You also say that KeepFlying helps asset owners protect the value and integrity of capital that moves between owners and across borders. Why is that so difficult and how does your product ease this burden?

Haran: This is currently difficult as in many cases, these data points between engineering and finance sit in silos and in many cases for good reason. We have seen lessors who’ve owned the PBH market like GECAS protect a lot of their models. I think with the type of digital tools and AI capabilities we have today, merging the finance and legal world with the airworthiness world is an essential step to ease this burden of running commercial and risk forecasts.

ATR: Your company talks about “enhancing airworthiness data integrity.” Why is this important and how does your product do that?

Haran: Given that airworthiness data exist between multiple data sources (ERP, M&E/MRO systems, scanned PDFs, Excels etc.), ultimately when trading or returning an asset from lease or maintenance, it is the paperwork that ascertains the value of the metal.

While this will change over time when more airlines and MROs have a “single source of truth”, at this moment in time, it is the scanned paperwork where the truth resides.

When dealing with large streams of data, asset specific models help identify anomalies using rules built specific to that asset type and operational conditions which enhances the airworthiness data integrity. This can save hundreds and thousands of man hours for records teams while also ascertaining integrity of data and analytics derived out of this.

ATR: You talk about data wrangling — please explain what this means and how it impacts assets.

Haran: The FinTwin interfaces with and draws data from the MRO / M&E System, as well as other digital sources of maintenance, airworthiness and operational data — mainly scanned PDFs.

AI/ML techniques are then applied to the data providing airlines, lessors and MRO facilities with the tools needed to generate aircraft, engine — or component — (asset) specific “what-if” simulations for specific future scenarios.

Given the array of formats and inconsistencies with which data is maintained across aircraft, engines and components, explainable AI is used to “wrangle” data for achieving integrity. The underlying KeepFlying platform is configured to “wrangle” data based on the type of asset being managed — including techniques to extract differences between task card revisions.

ATR: You also mention explainable AI. Is that different than AI?

Haran: It is a subset of AI. While building these data models, it is essential to understand the extended impact and biases inherent to the model to validate the assumptions and results being generated.

ATR: How are data wrangling and explainable AI related?

Haran: The rules that define the model when “wrangling” the data with speed and precision are those whose impact need to be understood and results interpreted.

For example, to be able to correlate what potential findings or scrap rates for vanes and blades can be discovered when say an HPT or LPT module is stripped and inspected even before the Engine enters the shop is trained using large data sets that span across engine age, stage length, environment, previous shop visit data, shop visit findings, EGT margin deterioration profile among others.

The model should be able to “explain” the weightages and biases against each factor that contribute to the eventual predictions.

ATR: Talk about data silos — why are these bad and how can your product break down silos?

Haran: Data silos will continue to remain a challenge as long as airlines and MROs run multiple systems to manage data across CAMO, Part 145, technical records & supply chain functions. Data wrangling and identification of the “single source of truth” data set is where KeepFlying can help break down silos and help avoid data redundancies and duplication.

ATR: Can KeepFlying help with the current supply chain challenges that are ongoing? How?

Haran: It will take another year or so for the supply chain challenges to ease. Having said that, the source for the material more or less remains the same for airlines and MROs. Some of our tools and capabilities around scrap predictions, for instance, can get you ahead of the queue. Accurate and advanced material planning can only ease the pain.

ATR: We are experiencing a shortage of qualified workers in the aviation maintenance workforce. How can KeepFlying help while this shortage exists?

Haran: The challenge post-pandemic is that a lot of valuable experience was lost as skilled personnel transitioned to other industries.

The average age of a mechanic, and I read this somewhere published I think by the FAA, is 51. For every four people leaving the industry, only three are joining back. Data platforms such as ours can certainly help bridge the transition with AI — for instance, generative AI applications in defect rectification can help save 15 minutes per defect. It certainly won’t be a replacement — no way, but can assist as new people enter the industry.