Building a Deep Tech Company? Most Startup Advice Doesn’t Apply — Read This Instead
Starting Up

Building a Deep Tech Company? Most Startup Advice Doesn’t Apply — Read This Instead

Learn from Loyal founder Celine Halioua’s rigorous approach to milestone-based planning, fundraising and multi-product strategy for deep tech companies.

“I tried to raise a Series B in Q4 of 2022, and it failed miserably,” Celine Halioua recounts with her signature frankness. As the solo-founder of Loyal (an animal health company developing the first drugs to help dogs live longer, healthier lives), she had the underlying scientific data, the big ambitious vision, and a compelling pitch — but as the saying goes, timing is everything. 

“The market turned. Then SVB happened. And nobody wanted to take the risk of telling LPs, ‘Sorry, we lost a shit ton of money on a dog drug.’ I ended up having to pivot to raising from angels and family offices, and it took me six months to raise $10 million, which is obviously terrible,” she says.

Fast forward just one year, when Loyal’s first dog longevity drug earned the FDA's first-ever formal acceptance that a drug can be effectively developed to extend lifespan for any species — including humans. With that simple one-page document, everything changed. “We got terms on our first day of our next Series B attempt after that,” she says. “It was probably the two most opposite raises you could have had.”

This stark contrast in fundraising fortunes is indicative not just of market timing, but of the challenges of building deep tech companies, where early progress isn't measured in users or revenue but in binary milestones that unlock — or destroy — value. 

“What’s funny is that we had the data that unlocked efficacy approval for a year and a half before we got it. I was very confident we were going to earn it, I just didn’t know when,” Halioua says. “But that didn’t count with growth-stage investors during our first raise attempt. While it was frustrating initially, when you step back, it makes sense. Proving efficacy is one of the most challenging parts of developing a novel drug. Checking off that section of our conditional approval application dramatically decreased the diligence load and the bet they’d be underwriting.”

Just a few weeks ago, Halioua was able to cross off another major milestone on Loyal’s journey by nabbing the second-ever efficacy approval for a longevity drug, this time for LOY-002, their chewable for senior dogs of almost all sizes. (She also announced a Series B-2 round, which brought total investment in Loyal to over $150M.)

As early backers of Halioua (since her 2020 seed round), we here at First Round could point to countless examples like these as to why she’s a founder who’s dangerous to bet against. But the rigor which she’s running her business with merits studying as well — in our view, her superpower lies in building an adaptable strategic plan and navigating a tricky milestone-based roadmap, relentlessly revisiting her priors along the way.

So in this exclusive interview, we’re diving into those topics, mining Halioua’s earned experience to surface the thorniest problems deep tech founders face, sharing the tactical solutions and frameworks she's developed to overcome them at Loyal. 

Should you focus on one approach or hedge with multiple products early on? How do you handle inevitable delays and puzzle through possible contingencies? How do you distill technically complex topics during fundraising and board meetings? And how do you maintain your motivation as a founder through a stream of setbacks? 

For deep tech founders navigating similarly long and uncertain paths to market, Halioua offers practical advice for navigating this terrain — where traditional startup advice often falls painfully short.

Deep tech challenge #1: The playbook paradox

From canine biology and federal regulations to drug manufacturing and design packaging, Loyal’s challenges don’t resemble those of B2B SaaS startups in the slightest, says Halioua.

“If you understand the consumer dog market, you almost certainly don't understand biotech. If you understand biotech, you don't get the veterinary market. And if you're a deep tech person, you probably haven't played in consumer. Generally, pharma sells to payers, and while Loyal’s drugs are prescribed by veterinarians, they’re going to be consumer-marketed at the end of the day — the ultimate customer is the dog owner advocating for their dog.”

This creates unique challenges, making Loyal a difficult company to evaluate for both biotech and consumer VCs alike. It also means that Halioua didn’t have many established playbooks to lean on.

For deep tech founders, it can be a company-killing move to just copy-paste advice for B2B SaaS. I've read what feels like every single piece of startup content out there, but it’s in order to educate my own model. Understand from first principles what applies to you, why, and what you’ll need to change to fit your business.

Solution: Lean on unexpected inspiration

Halioua recommends actively seeking out analogues, even if they're in seemingly unrelated industries. “A lot of what I've borrowed has been from companies like Boom,” she says. Her conversations with founder Blake Scholl helped shape her approach to scenario planning and milestone-based development. “Everything we do at Loyal is tied to milestones,” she says. 

“Certain hiring plans are allowed at milestones, certain capital spending is triggered by other milestones. We'll start working on this program when we get efficacy approval for that. It's the same for fundraising. I could go raise today at X situation, or I could raise at this milestone, and that increases the probability of that happening, but it adds risk in this new way.” 

For each milestone, Halioua develops multiple scenarios. “We have three scenarios until reality collapses on one of them at a particular milestone, and then we go down that path,” she says. “Before it was more of a vision pitch with great science behind it. Now I more intimately understand the levers that matter for our business.”

We have every single day mapped out on what it takes to go from today to approval, what can go wrong, and where we can slice things off if we need to.

Solution: Build a board that spans domains

“What makes this difficult is that people who could most help you navigate one domain likely have little understanding of the others you operate in. The solution is not to waste your time looking for a single unicorn advisor who understands everything, but to assemble more of a constellation of perspectives,” she says.

This is how Halioua has approached board composition. “I have a pretty big board — nine people, but not all voting. I had the choice between trying to find a founder who had experience in both veterinary medicine and biotech, or what I ended up doing, which was trying to find the best possible person in each vertical: veterinary regulation, deep tech operating, biotech company building, marketing,” she says. 

“For example, I added an FDA regulatory expert to our board. Now when we get an update from the FDA, I actually call on her first at the meeting and say, ‘What's your interpretation of this?’ so it helps everyone understand.” (And sitting in that deep tech seat? Scholl from Boom, of course.) 

Deep tech challenge #2: The delay domino effect

In deep tech, delays are inconvenient, if inevitable. “We had a surprise six-month study the FDA asked us for mid-last year, which we didn’t predict would happen,” she says. 

When you’re building in deep tech, you just get delayed in the weirdest of ways that you don’t expect. It humbles you.

Solution: Open another decision-making door

To navigate high-stakes decisions prone to delays with appropriate care, Halioua developed a framework that builds on Jeff Bezos's well-known one-way/two-way door concept: the slow two-way door.

“A lot of deep tech decisions fall into this bucket,” she says. “It’s technically reversible, but with big time or reputation or money costs associated.” Regulatory submissions are a perfect example. “When we submit something to the FDA, we’re locked into the compound and safety strategy, and we don't get feedback for six months.”

This framework creates clarity around how much time and attention different decisions deserve. “Even though it’s a two-way door and I push my team hard on moving fast, I will never push them on a final FDA submission to cut corners or submit faster. I will always take the extra week, sit on it, and feel confident.”

Solution: Rigorously scenario plan for inevitable delays

Halioua has since retooled Loyal to build around the expectation of the unexpected, with comprehensive contingency plans addressing every aspect of the business. “What do we still need to spend? What do we cut and when? What are our funding scenarios for if we're six months delayed to market, for if we’re 12 months late to market?”

She shares a specific example: “When we were waiting for the efficacy package, I sat down with my regulatory advisors, who've gotten dozens of drugs approved, and asked them, ‘If this goes wrong, what would that look like? Is there precedent for this happening? Is there precedent for that happening?’ Then we map out that we think there's a 40% chance of this and a 20% chance of that. And then we decide which scenarios are worth building redundancy against.”

Celine Halioua, founder and CEO of Loyal

Deep tech challenge #3: The fundraising slog

“Adoption, revenue, and other metrics typically derisk software startups and give early signs of product-market fit. You can find charts on LinkedIn that say, ‘At this ARR, you can raise a Series A; at this ARR, raise a Series B,’” Halioua says. 

“That doesn’t exist in deep tech, where derisking the company is predominantly tied to specific milestones that come with long wait periods. Us being five years in and about a year from market is pretty damn accelerated for getting a drug approved from 0 to 1,” Halioua says. 

Solution: Give investors new mental models and reframe the conversation around probability of success

Halioua learned the hard way that even when making scientific progress, investors might not value it without clear translation. “Typically the fundamental risk is whether people will pay for their product or whether the market is large enough. Our risk is, can we reach the market with an FDA approval? Can we find a drug that works?” she says.

“But sometimes I still meet investors who say, ‘I don't think the market's big enough.’ My reaction is always to say that there are a laundry list of risks to this company — TAM is not one of them,” says Halioua. “Apoquel, an anti-itch drug, makes about $800 million a year. The Farmer’s Dog has passed over $1B in revenue.  The COGS are great. We’ll have effectively 10 years of federally enforced exclusivity. There's market risk because nobody's approved a longevity drug before, but the baseline economics, even on day one, are going to be pretty good.”

While this disconnect was initially puzzling, Halioua has learned to take it in stride. “You have to create mental models for investors to evaluate a deep tech company. With traditional software, they know what to look for. But with a company like Loyal, we have to educate them on what matters and why,” she says.

Raising for a deep tech company is an art I've constantly had to hone and optimize. I was punched in the face a few times before I really developed an understanding of how to talk about what we're trying to do to a Silicon Valley audience.

This required Halioua to develop an entirely different fundraising framework — one built around communicating how they were inching closer to reaching FDA approval. “My advice is to build every fundraise around an increased probability of success of unlocking this market. Break down the drivers of that — both objectively, such as making regulatory progress, but also on what investors will care about,” she says.

“Now our fundraising strategy is based off of A) helping investors who don't normally invest in this space understand how to think about the market and what good should look like, and then B) understand where we are relative to it. So it's pretty formulaic: Here's where we are, here’s what we need for approval, here’s the differential, here's how we fail, here's how we succeed, here's why I think we're going to succeed, but here's the discount,” says Halioua.

For deep tech, the endpoint is usually objectively valuable. It’s not about convincing investors that people will want your product or that the business can scale — it’s about whether it’ll work (and if you can execute). When we shifted away from selling on vision alone and reframed in terms of our probability of success, it transformed our fundraising.

Solution: Expand your fundraising horizons

When traditional VC leads weren’t panning out for Loyal’s initial Series B attempt, Halioua got creative. “When ZIRP ended and everything was exploding, the sophisticated version of retail — high-net-worth individuals and family offices — saved us. We raised about $10 million in the summer of 2023 from that route, which is what gave us enough runway to go out and raise our big Series B from Bain later on.”

Though time-consuming, this experience gave Halioua confidence. “There’s a lot of money in this world and if you're willing to put in the ungodly hours, you can always raise the capital you need. You don't always have to pull from the same five, 10, 20 VC funds that everybody talks about,” she says. “If you're doing something weird or the market's weird, that can be good differentiation. People personally care about many of these deep tech ideas in a way that they don't care about a SaaS company.”

The key is to build the networks before you need it. “Family offices want to get to know you and deeply understand what you're doing,” she advises. “It’s a relationship, not a transactional process.”

Deep tech challenge #4: The binary bind

For software startups, iteration is king. Release an MVP, gather feedback, improve, and repeat. “For Loyal, due to the extensive regulation, the final product — the drug — is locked years before it launches. If the drug doesn’t work, there’s no iterating. If you want to modify the product, you need to restart the entire process over again,” she says.

“Regardless of your religious beliefs, there's a God in biology,” Halioua says. “When you’re attempting something that’s scientifically challenging, there's an objective truth that's already predetermined in the universe — and it's your job as a founder to figure out that truth as quickly as possible,” she says. “It's been predetermined since Day 0 whether LOY-001 will extend large dog lifespan. All I can do is make sure the experiment is run as cleanly and efficiently as possible.”

There's a chaos variable that you can't predict in deep tech. You can't force or iterate your way to making a drug if it ultimately doesn't work.

Solution: Build in redundancies, big and small

This immutable constraint led Halioua to take more of a portfolio approach of developing multiple products simultaneously, even when resources were scarce and conventional startup wisdom demanded focus. “A big debate we had in the early days of Loyal was why to spend the money to go multi-product so early. And I'm really glad we did — we'd be dead today if we hadn't,” she says.

She shares a specific example: “Our lead drug today is LOY-002, but we actually started working on it second," Halioua says. "We wanted to have redundancy in case something happened to LOY-001, something that we couldn't predict and didn't have any data to suggest would happen. Even though we felt the thesis was correct and it obviously ultimately got efficacy approval, we didn't want Loyal to fail on the chance that LOY-001 didn't work.”

Halioua’s thinking here applies to more granular decisions as well. “One example is our STAY study, which is our big, 1,000-dog lifespan and healthspan clinical study. It's a binary moment: We'll run the study for five years, and we'll either see statistically significant improvement in lifespan or we won't. But today, five years before that readout, we're doing things like thinking about year-four retention bonuses, the type of clinics we bring in, and doubling up enrollment to build redundancy. The goal is to create experiments where failure, if it happens, is clearly due to the approach itself rather than execution issues.”

When you're working on something deeply technical, there's not just your opportunity, there’s 20 opportunities around it. How do you not get distracted? But how do you still follow your intuition and get distracted just enough so you find new opportunities?

Solution: Engineer learning opportunities

Halioua’s determination to not place all of her product eggs in one basket has also led her to look for ways to remove blinders and allow for organic opportunities to emerge wherever she can.

“Early on, we decided to run what became known as The Healthspan Study, where we measured aging biomarkers in over 450 dogs old and young, big and small,” she says. “We knew we were going to have to run a large lifespan expansion study in dogs at some point and we'd never done it before, so we wanted to build the organizational muscle and learn by doing.”

The team went beyond the minimum requirements, investing additional resources that ultimately proved invaluable. “Since we had money at the time, we decided to biobank samples from these dogs. That data actually ended up becoming the foundational data to all of our FDA submission packages — but we couldn’t have known that at the time,” she says.

Focus is key, but opening yourself up for discovery is just as important when you're working on something deeply technical. You never know what information you’ll need. 

Of course, not all of these bets will pay off. “The counter example is a project we ran called X-Thousand Dogs,” she says. “We wanted to develop a way to determine a dog's biological age from a saliva swab. Essentially, we wanted to develop a way to create a health score for dogs.”

The project seemed promising, but didn't deliver proportional returns. “We ended up spending a ton of money on sending people these DNA kits for free, on getting the approach validated, on AWS infrastructure — and we haven't done anything with that data yet. It was multiples more expensive than the healthspan study, and we certainly learned multiples less. But it's always difficult to know that at the time,” Halioua says. 

Deep tech challenge #5: The founder’s emotional tightrope

Deep tech companies require the unbridled optimism needed to tackle impossible challenges, balanced with the calculated pessimism necessary to predict and hedge against frequent setbacks. This psychological balancing act becomes especially tricky over the years-long timeline. “You have to be so naively optimistic as a founder to be willing to do anything,” Halioua says. 

If I knew how hard building an FDA-approved longevity drug was five years ago, I probably wouldn't have even tried. But at later stages, you have to abandon optimism and actually be massively pessimistic.

Solution: Stay emotionally attached to mission, not method

Halioua's first guideline for this balancing act is to be unwavering about your mission while remaining flexible about how you achieve it.

“As you start making progress, you are forced to commit to a path. Now at this stage of the company, it's very regimented. But in the early days, it’s very open-ended. It’s ultimately about deeply understanding the vision that you’re emotionally attached to. What's the core bet of the organization? Everything else is just a strategy to achieve that,” says Halioua.

From Day 0, I have always been emotionally attached to getting the first FDA-approved drug for lifespan extension. I had a very strong hypothesis that the way to do this was to focus on big dogs with short lifespans. And early on when I was first starting as a solo-founder in 2019, I had a low conviction bet that the way to go about it was gene therapy to knock down IGF-1 production, which is what we think accelerates their aging,” she says.

This unwavering commitment to the core mission — rather than to a specific technical approach — allowed Halioua to pivot when necessary. “When I hired my scientific team and we were deciding on the right direction for LOY-001, I was crystal clear that I just want the highest probability shot on goal for getting drug proof for lifespan extension in dogs,” she says. “So they were comfortable pushing back on how the COGS on gene therapy would be terrible and safety would be tough. It turned out I was right about big dog, short lifespan, but I was wrong about gene therapy.”

Don’t get emotionally attached to your approach until there's objective data to prove that you should.

“I don't attach ego to literally anything but the end goal, so I don't give a shit if I'm wrong. I really, in fact, I assume I'm wrong,” Halioua explains with characteristic candor. “If I feel fire or if I have that niggling feeling in the back of my head, I run towards it because I'm not trying to protect a prior assumption I have. I’m trying to protect that end goal, and I'd much rather find out what's wrong today versus in two months.”

Deep tech challenge #6: The hammer-nail hiring trap

Deep tech companies require bringing on highly technical talent, but this specialization comes with a significant risk. Experts who’ve spent years becoming world-class in narrow domains often see every problem through the lens of their specialty.

“The hiring fail mode I often see is ‘hammer nail,’” Halioua says. “That's a reason a lot of biotech companies started by people straight out of their PhD fail. They're extremely emotionally attached — and have their blinders on—- for the problem they’ve spent 7+ years working on.”

Solution: Hire for thinking process over domain knowledge

Halioua has come to look for the ability to apply rigorous thinking to novel problems, rather than simply being experts in Loyal’s exact field.

"If you're doing something novel, there’s not a ton of specialized PhDs to even hire. Right now there’s only a couple of groups getting PhDs on dog longevity. So actually the background of your early scientific hires doesn’t matter in some ways. A good scientist can learn and ramp very quickly in a different field,” she says. 

“As an example, a lot of our early team had neuroscience backgrounds. We're not doing anything neuro related. But neuroscience is very complicated. A lot of things are not understood, and you have to go extremely deep to make any progress at all. So they're able to apply those mental frameworks to a new problem.”

Here’s how she hunts for this trait in interviews: “One of the biggest hacks is seeing the questions they ask you. With the best candidates, the quality of the question-asking is very clear. How much do they interrogate what you're doing and the rationale, and how do they then update their priors? 

When you’re trying to hire scientists, the best ones will assess you as much as you assess them.

She also tests depth of understanding. “I make them explain things to me. I'll pick some aspect of something they've worked on and go down deep, deep, deep. A lot of people will hit a level of depth and get stuck versus somebody who's great can go all the way to the atoms.”

Deep tech challenge #7: The technical translation gap

When you're building at the frontiers, developing trust with investors who lack the specialized knowledge to directly evaluate your work sets up another barrier. “There's such a trust aspect when you're doing something weird, hard and different. I used to be so pissed off about it,” Halioua admits. “Like, ‘Why do I have all these hoops to jump through?’”

Even once investors send the wire though, things don’t necessarily get easier. Running a top-notch board meeting is tough for any founder, but it can be especially challenging for deep-tech startups with heaps of technical info to communicate and complicated chess-strategies to consider. 

“I had somebody tell me recently that my first board meeting was the worst board meeting they'd ever been to,” Halioua laughs. “But I think at least my board meetings are at least competent now.”

Solution: Radical transparency

“Instead of complaining about our hurdles, I've tried to engineer it in a way where it's undeniable,” Halioua says. “Externalized signals of competency and of accuracy are super important here.

Halioua has also found that being extremely upfront about problems and mistakes is the fastest path to building trust. “Trust can be lost in an instant, so I'm always upfront about something going wrong,” she says. "When we got an FDA delay last year, I called Josh within an hour. I'm not sitting around coming up with some way to package it up. I default to, ‘This is what happened, and now we're figuring it out.’”

This approach extends beyond just sharing bad news to openly acknowledging mistakes. “When I’ve made bad hires, I've gone in front of my team and apologized. I've sent detailed emails to investors about firing an executive, and one of my investors was like, ‘Literally nobody does this. What are you doing?’ But my reasoning has always been that because I'm a solo founder building something extremely technically complex, I want to be ridiculously transparent about what I'm thinking and why.”

Solution: Apply a consistent formula

“It's your job to educate the board about what they need to know, but the lazy approach is just trying to teach them everything."

Here’s how she does it: “I now use this framework: Here's what we said we were going to do last time. Here's where we are today. Here's the delta. Here's why.” This consistent approach creates continuity between meetings. “That way I can start with, ‘Here are the five follow-ups from the last board meeting. Here's the status. We decided not to do item 2, here’s why. Here’s how we completed item 3, and here are the results.”

She’s also learned that board members need regular reminders of the overall framework, so for critical pathways like FDA approval timelines, Halioua provides clear tracking. “Here was the date I gave at the last board meeting. Here's the date today. What changed? Why? Which of these is the rate limiter? Which of these is high risk, medium risk, low risk?”

Don't make people remember things — that's something a lot of bio founders get wrong. It should be like a mantra: Here's the framework we’re using, here’s our progress, here’s the change, here’s the risk.

The deep tech advantage: turning challenges into moats

“There's a really good Sam Altman blog post on why doing hard things is paradoxically easier,” Halioua says. “I’ve felt that 100% with Loyal. There are so many problems that other companies have to deal with that we don't. We've had so many tailwinds. So many people want to join our company, we have an almost 100% offer acceptance rate. We’ve been on the cover of WIRED, The New York Times, The Wall Street Journal. The enthusiasm for what we’re doing is really high. People care.” 

Aside from the recruiting advantage, Halioua has come to spot other benefits as well. “There are many moats when building something deeply technical — patents, federal incentives, the team. A moat that is often overlooked is simply time,” she says.

“When working in atoms, it takes a fixed period of time to hit certain milestones. In animal drugs, every time you submit a technical section to the FDA the review period is six months. This is the same for a little company like Loyal and an $70B+ behemoth like Zoetis. No amount of resources can make those go faster — a valuable advantage for a startup. I have felt confident for years that this will eventually get approved. Really then the question is how long would it take?”

The milestone marathon continues, but now more than five years in, we’re getting closer to finding out.