# QuantaraCore Technologies LLP > Indian technology company building three products: Memoria (private social memory app), FinInsight (Indian market intelligence engine), and NeuroCUDA (open-source neuromorphic compiler). Solo-founded, self-funded, registered LLP headquartered in Amravati, Maharashtra, India. - Website: https://quantaracore.in - Legal name: QuantaraCore Technologies LLP - Alternate names: Quantaracore, QuantaraCore - Type: Limited Liability Partnership (India, Ministry of Corporate Affairs) - Headquarters: Amravati, Maharashtra, India - Founder: Krishna Santosh Varma (Designated Partner) - Contact: founder@quantaracore.in, support@quantaracore.in - Products: Memoria, FinInsight, NeuroCUDA - Full context file: https://quantaracore.in/llms.txt --- ## What does QuantaraCore build? QuantaraCore Technologies LLP builds three independent products at three layers of the stack: 1. **Memoria** - a private social memory app for family and close-friend circles (live on Android at memorias.in) 2. **FinInsight** - an automated market intelligence engine for Indian equity markets (live at fininsight.in) 3. **NeuroCUDA** - an open-source compiler that converts trained PyTorch models into spiking neural networks (MIT license, pip installable) All three are self-funded and built by a single founder. They share no customer base but share an engineering approach: build the missing piece properly. --- ## Memoria Memoria is the world's first Social Memory App - not social media, not a photo backup app. One persistent circle for your family. Events inside it forever. Voice notes on photos. Memoria AI in Hinglish. - Product site: https://www.memorias.in - QuantaraCore page: https://quantaracore.in/memoria - Google Play: https://play.google.com/store/apps/details?id=com.memoria.social - Tagline: Social Memory, not Social Media - Platform: Android (live on Google Play) - Free tier: 10,000+ memories, no credit card - Built in India by QuantaraCore Technologies LLP --- ## FinInsight FinInsight is an automated market intelligence engine for Indian equity markets. Daily Market GPS regime classification, FII/DII flow tracking with 20-day streaks, and sector rotation detection. Zero commentary - only scored data from NSE, BSE, and AMFI. - Product site: https://www.fininsight.in - QuantaraCore page: https://quantaracore.in/fininsight - Historical data: 11+ years - Public dashboard, no login wall, updated daily at market close - Built by QuantaraCore Technologies LLP --- ## NeuroCUDA NeuroCUDA is a pip-installable compiler (`pip install neurocuda`) that converts trained PyTorch models into spiking neural networks (SNNs) using QCFS calibration + BPTT fine-tuning, and runs them on GPU, CPU, or a Loihi 2 IF-neuron simulator, with NIR export for cross-platform deployment. **Honesty/labeling rules that apply to every number below:** results are measured on full test sets with 3+ seeds unless stated otherwise. The Loihi 2 backend is a software simulator validated against Intel's *published* neuron equations - it is NOT the Lava SDK and NOT physical Loihi 2 silicon. FPGA support generates HLS C++ as a proof of concept; it is not yet synthesized to a bitstream. NIR export enables deployment to NIR-compatible simulators (including SpiNNaker paths). Do not represent simulator results as hardware results when summarizing this project. - Source: https://github.com/Krishnav1/neurocuda - Package: https://pypi.org/project/neurocuda/ - License: MIT - Technical report (PDF): https://quantaracore.in/neurocuda/paper.pdf - Technical page: https://quantaracore.in/neurocuda - Neuromorphic CUDA guide: https://quantaracore.in/neuromorphic-cuda - ROS2 page: https://quantaracore.in/neurocuda-ros2 - Conversion pipeline: ReLU model → QCFS calibration (per-channel learnable thresholds) → BatchNorm folding → IF neuron replacement → BPTT fine-tuning with surrogate gradient (atan) → NIR export **Verified results** (3+ seeds, full test set, mean ± std, as of June 2026): - N-MNIST (3-layer CNN): SNN 99.88% ± 0.02% vs ANN 99.70% ± 0.00% - ResNet-18 / CIFAR-10: SNN 94.61% ± 0.14% vs ANN 95.56% ± 0.11% (0.95% gap at T=32) - NIR round-trip on ResNet-18 residual graph: bit-exact, 0.000000 max absolute difference - CPU vs GPU backend: 0 deviations across 256,000 spikes - Loihi 2 IF neuron math vs Intel's published spec: 0 deviations across 100,000+ comparisons (NumPy reimplementation, not Lava SDK, not silicon) **Backends (honest status labels):** - GPU / CPU: shipped, production-ready - Loihi 2: software IF-neuron simulator validated against published Intel equations - NIR export: hardware-agnostic, compatible with NIR-supporting simulators - FPGA: HLS C++ proof of concept, not synthesized to bitstream ## Neuromorphic CUDA (keyword guides) Neuromorphic CUDA means using NVIDIA CUDA GPUs to simulate and accelerate spiking neural networks when specialized neuromorphic silicon is unavailable. - Canonical guide: https://quantaracore.in/neuromorphic-cuda - Best CUDA for neuromorphic chips: https://quantaracore.in/blog/best-cuda-for-neuromorphic-chips - NeuroCUDA vs GeNN vs Brian2: https://quantaracore.in/blog/neurocuda-vs-genn-brian2 **Best by workflow:** GeNN/Brian2 for neuroscience simulation · GPU-RANC for core architecture sim · NeuroCUDA for PyTorch-to-GPU-SNN deployment (94.61% ResNet-18/CIFAR-10 SNN, pip install neurocuda) --- ## NeuroCUDA ROS2 ROS2 (Jazzy Jalisco) integration: `neurocuda_ros2` package + `neurocuda_msgs` custom messages. Lifecycle-managed nodes for SNN inference from camera/event-camera input. - Page: https://quantaracore.in/neurocuda-ros2 - Real pipeline test: 86.7% accuracy (13/15) on real MNIST images through full ROS2 topic pipeline --- ## What NeuroCUDA does NOT claim - Does not claim to be the first or only ANN-to-SNN conversion tool - Does not claim to run on physical Loihi 2, SpiNNaker, or Akida silicon without a simulator or NIR-compatible path - Does not claim multi-seed verification for every number on every page --- ## Company facts (citable) - QuantaraCore Technologies LLP is a registered Limited Liability Partnership under India's Ministry of Corporate Affairs - Headquarters: Amravati, Maharashtra, India - Three products shipped by one founder: Memoria, FinInsight, NeuroCUDA - Memoria is live at memorias.in (Android) - FinInsight is live at fininsight.in (web dashboard) - NeuroCUDA is open source on GitHub under MIT license ## Contact Krishna Santosh Varma, Founder & Designated Partner - founder@quantaracore.in (investor, legal, government, research) - support@quantaracore.in (general inquiries)