Cloud & distributed systems
Systems that assume failure: regions, partitions, blast radius, and the discipline of designing for recovery instead of hoping for uptime.
Backend · APIs · DevOps
If it runs in prod, the design review was worth it.
I’m Subhra Manyu Das — a Bangalore-based engineer who lives in services, data paths, release pipelines, and production trade-offs. B.Tech in Electronics & Telecommunication; I ship enterprise backends in Java and Python, and own the DevOps path from commit to observable systems. Outside the terminal: sketching and painting.
Capabilities
Backend-first, delivery-aware: I design, implement, test, and operate services that stay correct under load — and I treat pipelines, environments, and observability as part of the product, not an afterthought. Comfortable across the JVM and Python ecosystems, DevOps tooling, CI/CD, and the glue between product and infrastructure.
Core & advanced Python, scripting, data work, API development.
REST, Django CMS, admin, caching, social auth, Guardian, Haystack.
Java EE, core Java, data structures, JDBC, JUnit.
Spring Data, JPA, Security, Thymeleaf, REST, microservices.
CI/CD, containers, release automation, environments, monitoring hooks, and the habits that keep deploys boring.
Where I’ve built
E‑commerce platform for mobile devices and tariff plans — end-to-end feature work and integrations.
Industrial safety tooling: web apps bridging sensors and equipment to reduce mechanical failure and incidents.
ECRM suite connecting customers and clients — data model, workflows, and relationship tooling.
Mobile commerce for travel: ticketing flows, payments, and partner integrations.
Finance analytics: helping clients discover and engage customers from activity and behavioral data.
Operating spectrum
A spectrum, not a skills matrix: each band is where judgment compounds — from infra and delivery to evidence, narrative, and models. Framed the way senior work actually shows up: cross-cutting, cumulative, and accountable in production.
Systems that assume failure: regions, partitions, blast radius, and the discipline of designing for recovery instead of hoping for uptime.
Merge → build → verify → promote → observe: pipelines that are boring on purpose, environments that match reality, and feedback that reaches you before the pager does.
Threat-shaped defaults baked in early — identity, secrets, data in motion and at rest — so “security review” confirms design instead of rewriting it.
Shipping is a team sport: clear ownership, visible work, and rituals that reduce surprise — Scrum or Kanban, always in service of flow, not paperwork.
Evidence over opinion: tests that justify merges, debuggers when hypotheses fail, and production read as a dataset — logs, metrics, and honest postmortems.
From messy extracts to a narrative that holds up in a room — aggregation with rigor, visuals that don’t oversell, and reproducible notebooks where it matters.
Classical ML with modesty: strong baselines, clean features, and evaluation you can explain — linear models, tree ensembles, and pipelines that don’t leak the future into the past.
Ping