Most engineers pick one lane. I built a career at the intersection of five.

I design and ship full systems end-to-end — backend, frontend, infrastructure, security, and the AI layer on top — then I sit with the customer to make sure it survives production. That combination — builder + forward-deployed problem-solver + AI engineer — is why enterprise BFSI teams across APAC and North America have trusted me with their platform experience for 4+ years.

I don't hand off. I don't wait for a ticket. If a customer is blocked at 2 AM, I ship the fix, own the rollback plan, and write the postmortem the same week.

Senior Software Engineer based in Bangalore, India, currently building a secure Virtual Data Room platform for regulated enterprises. Previously Senior Solution Engineer at Vymo, where I owned CRM platform delivery for banks, insurers, and financial institutions across APAC and North America.

I care about three things, in this order:

1. Customer outcome — does the change actually move a metric that matters to the business?
2. System integrity — will it still be correct under load, failure, and adversarial input?
3. Team leverage — did I leave the codebase, docs, or on-call runbook better than I found it?

My comfort zone spans distributed systems, event-driven architectures, secure authorization (RBAC/ABAC), and LLM-powered product features — glued together with pragmatic engineering.

name:         Prabhakar Kumar
role:         Senior Software Engineer
location:     Bangalore, India
experience:   4+ years
domains:
  - Enterprise SaaS
  - BFSI / Fintech
  - CRM Platforms
  - Secure Document Rooms
  - Applied AI / LLM
stack:        polyglot full-stack
availability: open to select roles

I design systems the way I would want to inherit them: small services, explicit contracts, and no magic.

flowchart LR
    Client[Web Client] -->|HTTPS| APIGateway[API Gateway]
    APIGateway -->|JWT + ABAC| AuthZ[AuthZ Service]
    AuthZ -->|OK| DocSvc[Document Service]
    DocSvc -->|write| Mongo[(MongoDB)]
    DocSvc -->|emit| MQ((RabbitMQ))
    MQ --> Audit[Audit Service]
    MQ --> Replica[Cross-Region Replicator]
    MQ --> Search[Search Indexer]
    Audit --> AuditStore[(Immutable Audit Log)]
    Replica --> MongoDR[(MongoDB DR)]
    Search --> SearchIdx[(Search Index)]
    DocSvc --> Cache[(Redis Cache)]
    AuthZ --> Vault[[HashiCorp Vault]]

Security is not a feature I bolt on at the end — in BFSI and secure-data platforms it is the product.

I use LLMs the same way I use any other dependency — with an eval, a fallback, and a cost budget.

Virtual Data Room Platform Senior SWE · Security · Distributed Systems Enterprise-grade secure document collaboration. Owned RBAC/ABAC authorization, audit-logging framework, event-driven microservices, and cross-region MongoDB replication. Node.js RabbitMQ Kubernetes Helm Redis Vault MongoDB	Intelligent Lead Platform AI Engineering · Full-Stack AI-driven lead scoring, cohorting, and real-time allocation — routes leads across AI outreach, bot calling, and human sellers based on live signals. React Next.js TypeScript Python Node.js PostgreSQL MongoDB
CRM Automation Platform Full-Stack · Platform Engineering Data setup, validation, scheduling, and broadcast automation for enterprise CRM rollouts. Cut setup time by 80%. Node.js Express.js React MongoDB Microservices	Voice → Text Applied AI · Product Real-time voice-to-text pipeline with streaming transcription and speaker awareness. Built as an applied-AI exploration on top of streaming speech APIs and LLM post-processing. Python Streaming Speech APIs LLM Post-processing
AI Applications LLM · Agents · Prompt Engineering Internal AI tools — document Q&A copilots, agentic workflows on LangGraph, and CRM-side LLM assistants for enterprise reps. OpenAI LangChain LangGraph FastAPI Redis	More on GitHub Open source · Experiments Sandbox projects across TypeScript, Python, and infra tooling — the place I try things before they show up in production.

_{Primary Languages}

_{A static, always-rendering recognition strip — the classic trophy service is currently offline, so this is the durable alternative.}

• Publish a minimal, production-shaped RBAC + ABAC reference implementation in TypeScript.
• Ship a small LangGraph starter for enterprise document Q&A with eval baked in.
• Contribute upstream to a message-broker or observability project I use in production.
• Write one deeply-technical blog post per quarter — no listicles, no filler.
• Mentor at least two junior engineers into their first production incident (and out of it).

Title	Format	Status
Designing RBAC + ABAC for a Secure Document Platform	Blog post
Event-Driven Microservices: What I Wish I Knew on Day One	Blog post
Shipping LLM Features That Do Not Embarrass You in Production	Blog post
Forward-Deployed Engineering: The Underrated Career Path	Essay

Certification	Provider	Status
AWS Certified Solutions Architect — Associate	Amazon Web Services
Certified Kubernetes Application Developer (CKAD)	Cloud Native Computing Foundation
HashiCorp Certified: Vault Associate	HashiCorp
MongoDB Associate Developer	MongoDB University

• I have shipped code from three continents in a single quarter.
• My favorite debugging tool is still console.log — and I refuse to apologize for it.
• I read one systems paper a week. Recent favorites: DynamoDB, Kafka, and the Tigerbeetle design docs.
• I keep a personal on-call runbook — for my own life, not just my services.
• Non-negotiable morning routine: black coffee, one deep-work block, then meetings.

"I don't just write the code — I make sure it survives contact with a real customer." — my working definition of a Forward-Deployed Engineer

"Make it work, make it right, make it fast — in that order, and never skip the middle step." — Kent Beck

"The most dangerous phrase in production is: it worked on staging." — every SRE, eventually

Reach me on LinkedIn DM or email — read daily, replied to within 24 hours on weekdays.

Yes to Senior IC roles, AI/LLM product engineering, forward-deployed engineering, and technical solution engineering at high-trust teams.

Not looking for Cold recruiter blasts, unpaid "quick chats", or crypto-in-name-only roles. Everything else — always happy to talk.

_{Based in Bangalore, India · open to remote and hybrid · work-authorized in India}