It could do the reading for you.
Assistants, agents and retrieval systems that read, sort and draft, so your people decide instead of dig.
- 3 risks found
- Auto-renews, 60-day notice
- Liability capped at fees
- Data kept for 7 years
p.4p.12p.31
You’ve built the business. Let’s build what’s next.
What could your business do with the right technology?
It could exist by next month.
A working first version in weeks, on a foundation you won’t have to throw away when it takes off.
It could sell while you sleep.
Storefronts, platforms and internal tools that look like you and still work at two in the morning.
- New order#1042
- Payment received₹2,299
- Stock updated4 left
- Receipt sentemail
It could keep its secrets.
Audits, penetration tests and secure-by-design architecture, led by a Certified Ethical Hacker.
- signaturevalid
- issuer keymatch
- document hashchanged
Rejected · tampered
It could see Friday coming.
Pipelines, models and dashboards that turn what you already collect into what happens next.
MTWTF
forecastIt could prove it happened.
Smart contracts and ledgers for records that have to be trusted by people who don’t trust each other.
#10419f…2c
#10422c…a7
#1043a7…e1
#1044e1…5d
our word. See the work ↓
[ Selected work ]
Out in the world,
not in a deck.
Everything below is real: captures, code and marks from things we’ve built.
The Curate CrateStorefront, checkout & adminLive ↗
HeuriSightAI assessment platformIn production
NANDA ResolverVerified agent directoryLive ↗
NirvanaAlert management · identity & product
Oxide OSAgent-native microkernel, RustLive ↗
# pip install inscrape from inscrape import Inscrape client = Inscrape("sk_your_token_here") # Scrape any URL result = client.scrape("https://example.com") print(result.content) # Get Markdown md = client.markdown("https://example.com")
InscrapePython SDKOn PyPI ↗
HeuriSightCompetency profile
$ npm i -g @gkganesh12/codecraft-cli $ codecraft init $ codecraft › /plan scope the change › /code write inside the guard › /verify check it against the rules
CodeCraft CLIAI governance for dev teamsOn npm
CookProof-of-work cards for buildersIn build
$ brew install gkganesh12/tap/gk-stack $ gk-stack init $ claude -p "/ship" Tests 1 failed | 4579 passed › Stopping here. I won’t commit or push broken code.
GK-StackGuardrails for AI-assisted developmentOn GitHub ↗
SignetEU Cyber Resilience Act scannerIn build
# POST /api/v1/documents
{
"evaluator": "citation_existence",
"severity": "CRITICAL",
"confidence": 0.95,
"explanation": "Citation '123 F.3d 456
(9th Cir. 2019)' does not match any
opinion in CourtListener."
}
CiteGuardCitation checks for AI-written legal draftsOn GitHub ↗
CyberSentinelSecurity operations dashboardDemo ↗
BaseraStudent housing finder, PuneConcept
you scan 10.10.10.5 and find vulnerabilities
agent runs nmap -sV -sC
→ finds Apache 2.4.49
→ looks up CVE-2021-41773 (CVSS 9.8)
→ stores finding in knowledge graph
Oh-My-MachineAI console for authorised security testing
HemoLedgerBlood-bank traceability on Hyperledger FabricOn GitHub ↗
from model_train import PrivacyPreservingCTGAN # CTGAN trained with DP-SGD (Opacus) model = PrivacyPreservingCTGAN( use_dp=True, epsilon=1.0, delta=1e-5 ) metrics = model.train(df) print(metrics["epsilon"]) # privacy spent
Synthetic Data GeneratorPrivate tabular data, differential privacyOn GitHub ↗
Built to be used, not just shown.
The Uncovering · studio still-life series
[ How we build ]
Every surface has an underneath.
Move your cursor. The lens shows what’s running behind each screen. We design the part you see and engineer the part you don’t.
Surface 01 · The Curate Crate
A gift shop that feels hand-wrapped.
Underneath: products, stock, orders and accounts modelled so nothing gets lost between the cart and the courier.
model Product {
id String @id
name String
slug String @unique
priceInr Int
mrpInr Int?
category String
occasions String
featured Boolean
stock Int?
orderItems OrderItem[]
reviews Review[]
}
Surface 02 · HeuriSight
A dashboard a student can read in one glance.
Underneath: two engines, retrieval and heuristics, mapping every conversation to a competency framework.
- Conversationstudent work comes in
- Retrievevector store · documents
- Reasonknowledge graph · heuristics
- Map10-category competency model
- Showdashboard · cohort view
Surface 03 · NANDA Resolver
A handle you can type. An answer you can trust.
Underneath: every record is signed, canonicalised and checked. A tampered document is rejected, on screen, in real time.
def verify_agent_facts(doc, key):
proof = doc.get("proof")
if proof["verificationKey"] != key:
raise VerificationError(…)
body = copy.deepcopy(doc)
body.pop("proof", None)
try:
VerifyKey(pub).verify(
_digest(body), sig)
except BadSignatureError:
raise VerificationError(…)
[ What we bring ]
- Brand & interface design
- AI systems
- Full-stack engineering
- Security review
- Data & ML
[ What we believe ]
DesignEngineering
earns
attention.trust.
Most studios are good at one. You need both in the same room, arguing early, so the thing that looks right also works right. That’s the whole studio.
One last thing, about the name.
Aletheia — truth, with nothing hidden.
a·le·thei·a / ἀλήθεια /
noun, Greek. Truth. Literally “un-hidden”: the state of a thing when nothing about it is concealed.
It’s how we work. You see what we’re building, what it costs and what’s left, every week, without having to ask.