QuantaraCore Technologies LLP

We build the things people use every day, and the infrastructure most companies never bother to build at all.

QuantaraCore Technologies LLP is a registered Indian technology company building Memoria (private social memory app at memorias.in), FinInsight (automated Indian market intelligence at fininsight.in), and NeuroCUDA (open-source PyTorch-to-SNN compiler). One founder, three products, all self-funded and shipped from Amravati, Maharashtra.

3
Products shipped
1
Founder, all three products
94.61%
NeuroCUDA SNN accuracy (ResNet-18, 3 seeds)
11+ yrs
FinInsight historical market data

Three products. Three layers of the stack.

Social

Memoria - Social Memory App for Families

One circle for your family. Events inside it forever.

LIVE · memorias.in · Google Play

The world's first Social Memory App - not social media, not a photo backup. Private circles, structured events, voice notes on photos, and Memoria AI in Hinglish. No public feed, no ads, no algorithm. 10,000+ memories free on Android.

  • Persistent family circles - one home for every Diwali, trip, and reunion
  • Events inside circles keep photos and voice notes organized forever
  • Memoria AI search in Hinglish: "Goa trip with Rahul" or "dadi ke ghar ki photos"
  • Invite-only, end-to-end encrypted, zero ads and zero algorithmic ranking

Intelligence

FinInsight - Indian Market Intelligence Engine

Market GPS, FII/DII flows, sector rotation - zero commentary

LIVE · fininsight.in

Automated engine that reads NSE, BSE, and AMFI data daily and outputs a classified market regime with confidence score, institutional flow direction, and sector rotation signals. No opinions, no narrative - just scored, dated data on a public dashboard.

  • Market GPS: daily regime classification with numeric confidence score
  • FII/DII flow tracking with 20-day cumulative streak detection
  • Sector rotation across every NSE-listed sector
  • 11+ years of historical data, updated daily at market close, no login wall

Infrastructure

NeuroCUDA - Multi-Target Neuromorphic Compiler

One line of PyTorch. Any neuromorphic chip.

OPEN SOURCE · TECHNICAL REPORT

Convert a standard PyTorch model into a spiking neural network and run it on GPU, CPU, Loihi 2 simulator, or physical SpiNNaker-1 silicon (EBRAINS jobs #420148 and #420186 confirmed SUCCESS on two boards). NIR export for portable graphs. Backends are honestly labeled: SpiNNaker = real silicon; Loihi 2 = simulator only; GPU/CPU = published accuracy.

pip install neurocuda
  • SpiNNaker physical silicon confirmed - two independent Manchester boards, July 2026
  • ResNet-18 on CIFAR-10: 94.61% ± 0.14% SNN accuracy on GPU (0.95% gap from ANN at T=32)
  • NIR round-trip is bit-exact on residual network architectures (0.000000 max difference)
  • GPU and CPU backends shipped; Loihi 2 remains a validated simulator (not Loihi silicon)
  • 93.67% activation sparsity measured on NeuroBench
  • ROS2 package: run a compiled SNN on a robot's camera feed with one launch file

Neuromorphic computing & compiler resources

Long-form guides from QuantaraCore on neuromorphic CUDA, PyTorch-to-SNN conversion, Intel Lava alternatives, and NIR. Every page links primary sources.

SpiNNaker physical silicon confirmed (jobs #420148, #420186) NeuroCUDA - PyTorch to spiking neural network compiler Neuromorphic CUDA: complete guide (GeNN, Brian2, GPU-RANC, NeuroCUDA) Best CUDA for neuromorphic chips (2026 ranking) Intel Lava archived: what to use instead Loihi 2 + PyTorch without Lava (2026) What is NeuroCUDA? Open SNN compiler explained How to convert PyTorch to SNN pip install neurocuda: complete install guide What is NIR? Neuromorphic Intermediate Representation explained ANN to SNN conversion tools compared Best compilers for neuromorphic chips in 2026 FinInsight - Indian market intelligence engine All blog posts →

One company.
Three different products.

QuantaraCore Technologies LLP is a registered Limited Liability Partnership under the Ministry of Corporate Affairs, Government of India. Rather than building one product for one market, the work spans three distinct layers: a consumer app people use daily, a market intelligence engine professionals check every morning, and a compiler that other engineers can build on.

Memoria is the social layer: a private memory-sharing app for circles instead of public feeds. FinInsight is the intelligence layer: automated market regime classification for Indian equities. NeuroCUDA is the infrastructure layer: an open-source compiler for neuromorphic hardware. Each one is independently live or in active research, and each is built by the same person.

Solo-founded, self-funded, and headquartered in Amravati, Maharashtra. Read the full story behind that decision →

Common questions

What does QuantaraCore Technologies LLP actually build?

Three products at three different layers: Memoria, a private social app for sharing photos and memories within small, trusted circles; FinInsight, an automated market intelligence engine that classifies Indian equity market regimes daily; and NeuroCUDA, an open-source compiler that converts PyTorch models to SNNs - with confirmed physical SpiNNaker-1 silicon via EBRAINS and a Loihi 2 simulator backend.

Is QuantaraCore a registered company, or just a side project?

QuantaraCore Technologies LLP is a registered Limited Liability Partnership under India's Ministry of Corporate Affairs, headquartered in Amravati, Maharashtra. It has a designated partner, a legal address, and formal contact points for investor, legal, and government correspondence - see the Company section below.

Who builds and funds QuantaraCore's products?

All three products - Memoria, FinInsight, and NeuroCUDA - are self-funded and built by a single founder. Each product ships and operates independently of the others.

Why is QuantaraCore based in Amravati instead of Bangalore or Hyderabad?

Because the founder lives there, and because deep technical work - writing compilers, building financial data pipelines - benefits more from long, distraction-free focus than from proximity to a startup ecosystem. The full reasoning, including the tradeoffs, is covered in our blog.

Who founded QuantaraCore?

QuantaraCore was founded by Krishna Santosh Varma, who is the sole founder, designated partner, and builder of Memoria, FinInsight, and NeuroCUDA. All three products are self-funded.

What is the relationship between QuantaraCore and Memoria?

Memoria is a product built by QuantaraCore Technologies LLP. The app is live at memorias.in. QuantaraCore is the parent company; Memoria is the consumer social-memory layer.

Is NeuroCUDA open source, and can I use it today?

Yes. NeuroCUDA is MIT licensed and pip-installable (pip install neurocuda). Source is on GitHub. The technical report is at quantaracore.in/neurocuda/paper.pdf.

QuantaraCore Technologies LLP
Limited Liability Partnership
Maharashtra, India
Amravati, Maharashtra
Ministry of Corporate Affairs
Memoria, FinInsight, NeuroCUDA

Get in touch.

Whether you are an investor, a researcher, a potential partner, or someone who wants to learn more - we would love to hear from you.

Founder & Designated Partner

Krishna Santosh Varma

For communication regarding DPIIT recognition, Startup India, banking, investor relations, government matters, or research collaborations.

Amravati, Maharashtra, India