We separate the science from the science fiction.
Independent technical due diligence on deep tech deals, led by PhD-level industry experts in the exact technology you are looking at. A first read within 24 hours, a full report inside two weeks, and every finding tied to what it means for the investment.
Photonic interconnect startup, Series A
Key-person dependency on a single photonics PhD; no documented process knowledge | ||
Bench results at 4 nodes; no evidence of scaling beyond lab conditions | ||
Core patent claims overlap 2019 prior art; freedom to operate not assessed | ||
No load test beyond the current user base; single-region deployment, no recovery plan |
combined experience
ETH Zürich
EPFL
Imperial College
Google
IBM
World Economic Forum
Unit 8200
High-Tech Gründerfonds
EquityPitcher
PostFinance
Kodori
PwC
Swiss SECO
QBIT Capital
Lightbird
AcormaWorld Economic ForumDeep tech is a black box, until it isn't.
Groundbreaking science and sophisticated science fiction look identical from the outside. Here is one Series A deal, five claims from the deck, and what we found.
| The pitch deck says | What we found |
|---|---|
| “TRL 8, ready for pilot deployment.” | TRL 6. Bench demonstration at four nodes, no field data, coherence not measured at operating scale. |
| “Patented core architecture.” | Two patents granted. Independent claims overlap 2019 prior art; freedom to operate never assessed. |
| “A team of twelve engineers.” | Core photonics knowledge sits with one PhD. No documented process; her departure would halt the roadmap. |
| “Cloud-native, scales linearly.” | No load test beyond the current user base. Single-region deployment with no recovery plan. |
| “SOC 2 in progress.” | No penetration test on record. API keys committed to repository history, retrievable after deletion. |
Have a deal like this on the table?
Send the deck. Julien reads it personally and tells you within one business day whether a full assessment is worth it, and what it would cost.
Technology, business, and people.
A technical due diligence that only looks at code misses the full picture. We evaluate all three dimensions that determine whether a deep tech startup will actually succeed, and the report places the deal inside the triangle so your committee sees at a glance where the risk sits.
Technology
What the methods, code, data, and results actually mean.
- Is the approach defensible?
- Is the architecture scalable?
- Do the claims hold up under independent scrutiny?
Business
How the technology can solve real problems and create commercial value.
- Does it address the stated use case?
- Could it be applied elsewhere or sold as separate products?
People
Skillset, team cohesion, and transparency.
- Have key people left?
- Is the team capable of executing the roadmap?
- Are they being straight with you?
Defensibility
How difficult is it to replicate? Is there a real moat, or is this an off-the-shelf approach with a good story?
Scalability
How well will the technology follow growth? Will the architecture hold under production load?
Claims
Do experiments reflect real-world performance? Does the code match what was presented? Are benchmarks rigorous?
Legal & Financial
We work alongside your legal and financial due diligence teams for complete, integrated coverage.
Positions are illustrative · Orbits show that each approach stays inside its quadrant
Rigour and context, in the same room.
A lab tells you whether the science is possible. An auditor tells you whether the process was followed. Neither tells you whether this company can turn the science into the product in the deck. We do both, because our experts have done both.
- University and research labsFeasibility only
- Generalist IT consultantsNo domain depth
- Big 4 and in-house M&AProcess, not code
- Deep Tech ExpertsBoth, in two weeks
The same method, at four depths.
Pre-seed to M&A. Approach, timeline and deliverables match where the deal is, so a same-day read is possible when that is all the round allows.
A domain expert, not a generalist.
Every audit is led by someone who has published in, built in, or led teams within the exact technology.
Thirty-six ways a deal goes wrong.
Every risk we have logged across a hundred diligences, mapped by domain and severity. The closer to the centre, the more often it kills a deal.
The full register · 36 findings · eight domainsRead every finding as text.Each risk with its regulatory anchor, plus how we turn findings into a recommendation.
AiArtificial intelligence
Regulatory exposure & technical safety gaps
- EU AI Act classificationsystem misclassified as limited-risk when it meets high-risk criteria under Annex III; triggers full conformity assessment obligations
- Explainability gapsArticle 13 transparency requirements not met; black-box models in regulated sectors (credit, hiring, healthcare) face immediate compliance blocks
- Training data provenanceundisclosed use of scraped or licensed data; GDPR and copyright liability unquantified
- Jailbreak surfaceLLM-based products with no adversarial testing; prompt injection, goal hijacking, and output manipulation not evaluated
- Model driftno monitoring pipeline in production; accuracy degradation post-deployment undetected
Critical when present · Very high frequency in AI deals
CyCybersecurity
Vulnerabilities that invalidate the valuation
- No penetration test on recordor last test >18 months ago; unvetted attack surface in a product handling customer data
- Hardcoded credentialsAPI keys, passwords, and tokens committed to version control history; often retrievable even after deletion
- CVE backlogknown critical vulnerabilities in production dependencies unpatched; CVSS ≥7.0 items unaddressed for >90 days
- No incident response planGDPR Article 33 requires 72-hour breach notification; no documented procedure means regulatory penalty is near-certain post-incident
- Single-factor admin accesscloud consoles, CI/CD pipelines, and production databases accessible without MFA
Critical · Unpatched vulnerabilities can void cyber insurance and trigger GDPR breach liability
QcQuantum computing
TRL inflation and roadmap dependencies
- TRL inflationlab demonstration at 5–10 qubits marketed as a scalable product; coherence times and gate fidelities not disclosed at operating scale
- Error correction overhead not modelledlogical qubit counts assume perfect hardware; fault-tolerant qubit cost (1000:1 physical-to-logical ratio) not in financial projections
- Quantum advantage unprovenclaimed speedup not demonstrated against best classical algorithm on problem-relevant input sizes
- Hardware supplier dependencyalgorithm layer dependent on specific hardware vendor (IBM, IonQ, etc.) with no commercial SLA for qubit access at scale
High frequency · TRL gap is the defining risk in quantum deals
EnEncryption
Deprecated standards and post-quantum exposure
- Deprecated algorithms in productionMD5, SHA-1, RSA-1024, or 3DES still active in authentication or data-at-rest; known-broken and no migration plan
- Key management gapssymmetric keys hardcoded or stored in environment variables; no HSM or secrets manager in use; rotation policy absent
- Post-quantum migration not scopedNIST PQC standards (ML-KEM, ML-DSA) finalised 2024; no assessment of cryptographic asset inventory or migration timeline
- FIPS 140-2/3 compliance gapUS government and financial sector customers require FIPS validation; non-compliant modules block enterprise sales
Medium frequency · Critical severity when present in regulated sectors
AeAerospace
Safety, certification, and supply chain exposure
- DO-178C / DO-254 DAL mismatchsoftware or hardware design assurance level not aligned with actual failure consequence; re-certification cost can exceed development cost
- Single point of failure in safety pathno redundancy architecture for critical functions; fails airworthiness standards for commercial operation
- EMC / EMI not testedelectromagnetic compatibility testing not completed; product cannot legally operate in most jurisdictions
- ITAR / EAR export controltechnology with defence application not classified under US export control regulations; international sales blocked, investor returns constrained
- Fab concentration risksole-sourced from single foundry (commonly TSMC N5/N3); no second-source strategy; lead time >52 weeks with no mitigation
Critical · Certification gaps can halt commercialisation entirely
RoRobotics
Safety certification and real-world performance gaps
- Sim-to-real gap undisclosedperformance benchmarks from simulation not replicated in uncontrolled real-world environments; deployment readiness overstated
- Functional safety not certifiedISO 26262 (automotive), IEC 61508, or ISO 13849 compliance not achieved; product cannot operate in target sector
- Edge case coverageperception and decision systems not stress-tested on distribution-shift inputs; long-tail failure modes uncharacterised
- Liability framework absentno insurance, no product liability structure, no incident reporting procedure; one adverse event can terminate operations
Critical · Safety gaps are binary go/no-go blockers
SySystems architecture
Architecture debt and operational fragility
- No load testing evidenceclaimed capacity figures not validated under simulated peak load; architecture untested beyond current user base
- No disaster recovery planRTO and RPO not defined; backup strategy untested; single-region deployment for multi-tenant SaaS
- Vendor lock-inproprietary database, single-cloud dependencies, or third-party APIs with no abstraction layer; switching cost not disclosed in financials
- Technical debt ratioestimated remediation cost >20% of codebase undisclosed; post-acquisition refactor cost materially affects returns
- Missing observabilityno structured logging, distributed tracing, or alerting in production; incidents discovered by customers, not engineering
High frequency · Scalability gaps amplify with growth
SeSemiconductors
Supply concentration and IP hygiene
- Fab concentration risksole-sourced from single foundry (commonly TSMC N5/N3); no second-source strategy; lead time >52 weeks with no mitigation
- Yield data not disclosedwafer yield and binning data withheld, or reported from engineering lots only; unit economics at volume unverifiable
- Open-source IP contaminationcopyleft-licensed RTL, IP blocks or EDA scripts embedded in the proprietary design via transitive dependency; can force disclosure of design files
- Untracked dependency treeno software composition analysis (SCA) tooling; licence obligations for hundreds of libraries unknown
High severity when present · Foundry lead times set the timeline
For each risk identified, we assess its severity in the context of the specific deal — distinguishing a hard red flag that warrants walking away from a manageable issue that can be mitigated with targeted fixes, contractual protections, or post-investment remediation. Every finding comes with a clear recommendation: block, negotiate, or remedy.
Identifying these risks is exactly our job. Tell us about your deal and we will tell you what to look for.
Brief us on a deal →Led by the people who published the papers, then built the products.
One principal reads every deal and matches the expert. You always know who is doing the work, and you can attend every call.
Words from the people we have worked with.
A full suite of deep tech services.
Technical due diligence is our core offering, but our experts cover a wider range of needs for investors, corporates and technology teams. Like an X-ray machine, we see the details that matter, whatever the decision in front of you.
Workshops & Masterclasses
Expert-led sessions that give investment teams, boards, and executives genuine working knowledge of deep tech — so they can ask sharper questions and make better decisions.
Specialised Audits
Targeted assessments for specific technical risks that go beyond a standard due diligence — for when you need depth in one dimension, not breadth across three.
Vertical Specialisms
Domain-specific assessments for sectors where regulatory complexity, clinical validity, and technical risk intersect in ways that general technology due diligence cannot fully address.
Everything you need to know.
Still unsure? Reach out directly — we are happy to scope your engagement before you commit to anything.
Send an email →What is technical due diligence and why does it matter for VC?
Technical due diligence is a structured expert assessment of a startup's technology — covering IP defensibility, architecture, scalability, team capability, and the validity of technical claims. For deep tech investments it is essential: groundbreaking science and sophisticated science fiction look deceptively similar, and a single oversight can wipe out an investment.
How long does a typical engagement take?
Pre-seed engagements can be completed same-day. Seed to Series A typically take two weeks. Growth-stage assessments run three weeks. M&A engagements are scoped individually. If you have a hard deadline, tell us — we accommodate urgent timelines.
What does the deliverable look like?
You receive a structured SWOT report covering defensibility, scalability, and claims verification, plus a competitive benchmark. For technical audits, the report adds a prioritised action roadmap. We also debrief you after each call. You are welcome to attend all calls with the startup team.
How do you match the right expert to our deal?
When you brief us, we review the technology and identify an expert with direct hands-on experience in that specific field — not adjacent knowledge. For deals spanning multiple domains, we deploy a multi-expert team.
Can you work with the startup itself, not just investors?
Yes. Our Technical Audit service is designed for startups and their investors who want a concrete roadmap to make the technology more defensible and scalable. It delivers the same SWOT analysis as a due diligence, plus specific action points.
Do you handle NDAs and confidentiality?
Absolutely. All engagements are governed by an NDA before any information is shared. Our domain experts are bound by the same confidentiality obligations. We have extensive experience operating within strict VC confidentiality requirements.
What are the Masterclasses?
We offer expert-led masterclasses on AI, Cybersecurity, and Quantum Computing to help investment teams, boards, and LP audiences build genuine working knowledge of deep tech. If you need your team to ask smarter questions of founders, a masterclass with one of our domain experts is the most efficient path.
Brief us on a deal.
Two paragraphs are enough. Julien replies within one business day with a view on whether an assessment is worth it, the expert he would match, a timeline and a price. No commitment.
- Initial responseWithin 24h
- Expert matchedDay 1–2
- Seed–A report~2 weeks
- Growth report~3 weeks
Investors we have worked for will take your call. Ask, and we connect you within a day.