Systems architecture
We design large systems from a blank page: domain modeling, service boundaries, data flow, failure modes. Architecture meant to be operated, not just demoed.
Software architecture · Real-time systems
Pashupata Consulting is a software studio for the hard parts — real-time, fault-tolerant systems like algorithmic trading engines, carrier-grade communications platforms, and AI-driven products. We own the architecture, the implementation, and the part where it has to keep running.
Capabilities
Six things we do deeply — not a service menu, the actual hard problems we keep solving in production.
We design large systems from a blank page: domain modeling, service boundaries, data flow, failure modes. Architecture meant to be operated, not just demoed.
Low-latency pipelines for market data, telephony media, and live streams — with back-pressure, idempotency, and reconciliation built in, not bolted on.
Broker and exchange integrations, risk engines, order-state machines, fund-unit accounting, and live / paper / shadow execution that proves itself before it touches money.
Carrier-grade voice and messaging: SIP, RTP, and SMPP, real-time media and session-border control, omnichannel routing, multi-tenant contact centers, and SMS gateways at throughput.
LLM-curated decisioning, news and sentiment pipelines, and automation agents — wired into real products with guardrails and human review, not demos.
FastAPI and React / Next.js through to Docker, Postgres, Redis, and Cloudflare. We ship the whole thing — and we keep it running.
Selected work
Most of this is live in production right now — designed, built, and operated in‑house. It's the same depth of engineering we bring to client work.
Intraday NSE F&O auto-trader. Thirty-five independent strategies on index options and futures, hard-gated trading windows, and broker-side bracket orders backed by a Python reaper — belt and suspenders on every fill.
Intraday equity auto-trader running two signal engines in parallel across 209 NSE F&O stocks — an Astro-Neural model (XGBoost over five years of price data plus Vedic-ephemeris features) and a conventional technical engine — each routed through a SHADOW / PAPER / LIVE execution pipeline on Kite Connect, behind layered risk gates and kill-switches.
A multi-tenant Contact-Center-as-a-Service platform on a carrier-grade real-time telephony control plane — session-border control, media servers, IVR, recording, and skills-based queueing — unifying voice, WhatsApp, SMS, web chat, email, and social behind a per-feature, per-seat licensing engine.
A white-label crypto platform on Binance with pooled, mutual-fund-style unit accounting: daily NAV, per-user high-water marks, and automated weekly settlement — resellers never compute a distribution by hand.
A complete gamified learning platform for grades 6–12: story-driven interactive games across maths and science, a live exam module and question bank, classroom and lesson delivery, and linked student, parent, teacher, and admin portals — with leaderboards, progress analytics, credits, and payments on a real-time Firebase backbone.
A media-intelligence and public-accountability platform: it ingests speeches, interviews, and broadcasts, transcribes them by speaker, and surfaces contradictions and position-reversals over time. Every claim is backed by timestamped evidence and compiled into defensible, statute-referenced proof through a reviewed, fully audit-logged workflow — multi-tenant, role-based, with distributed processing.
Also in production & in the lab
How we work
The mark we carry is a trishul. Its three points are the only process we believe in — and we don't hand off between them.
We start at the architecture — the domain, the data, the failure modes — and design a system meant to be operated, not just to pass a demo.
We implement the whole thing — backend, frontend, infrastructure — in tight iterations, with the tests and reconciliation that real money and real calls demand.
We deploy it, watch it, and keep it alive. Most of our work is live in production right now. We build software we're willing to be on call for.
Work with us
The same engineering we put into our own platforms — applied to yours. Three ways teams engage us.
Greenfield systems designed and built end to end — backend, frontend, infrastructure — and operated in production, the way we run our own platforms. Real-time, fault-tolerant, and built to scale.
System design, architecture reviews, and the hard technical calls for teams building real-time, high-stakes, or just genuinely complex software — from data models and service boundaries to failure modes.
Profile, harden, and re-architect systems that have outgrown their original design — cutting latency, lifting reliability, and bringing infrastructure cost back under control.
Public sector & regulated
The controls government, PSU, and regulated clients need — designed in from the start, not bolted on.
We build and operate our own infrastructure end to end — so we can deploy on yours: on-premise, sovereign cloud, or air-gapped. Your data stays where policy requires it.
Immutable audit logging, role-based access, and defensible, statute-referenced records — the same model that powers FlipTracker.
Encryption at rest, hardened secret vaults, and least-privilege access across every system we ship.
We run production systems every day. You get a partner who stays on call and accountable — not a contractor who ships once and disappears.
The toolkit
Chosen because it survives production — and because we know it cold.
Get in touch
Tell us what you're building. If it's real-time, high-stakes, or just architecturally hard, it's exactly the kind of problem we like.