Sarim Ahmed.
Résumé

Senior Software Engineer at MiraBengaluru, India

I build scalable backends, distributed systems and production LLM agents, plus the AWS infrastructure under them.

RésuméEmail
users on Klok, the RAG product I built and scaled
300k+
platforms unified into one event layer at Mira
20+
request volume on the distributed LLM routing platform I architected
5×
GitHub stars on PennyWise, my open-source app
550+

Experience

3y 8m full-time

  1. Sep 2023 - Present

    • Led the rebuild of an AI customer-support agent on Agno: a 20+ tool registry, Milvus-backed memory, scheduled actions, and guardrails with auto-escalation for resilience against model variability.
    • Own the RAG and evaluation pipeline: a retriever service, multi-source ingestion with semantic chunking and deduplication, a multi-stage classifier, and a BAML eval service that reduces hallucinations on benchmark datasets.
    • Built an integration layer connecting 20+ platforms (Slack, Gmail, Teams, Zendesk, HubSpot, Jira, Linear, Notion, GitHub and more) into one event schema over SNS and SQS, with OAuth, dead-letter replay and per-tenant isolation.
    • Architected the Mira Network, a distributed LLM routing platform (FastAPI router, TypeScript nodes, Go CLI), scaling request volume 5×. Built a YAML workflow SDK and CLI adopted by 100+ developers.
    • Built and scaled Klok, a crypto RAG product, to 300,000+ users with 1,000+ concurrent daily, on FastAPI with AWS RDS and Fargate.
    • Tech-led a pod of junior engineers, interviewed candidates, set code-review standards, ran on-call, and owned the AWS infrastructure (Pulumi, ECS, RDS, IAM) and the Grafana observability stack.

    Python, TypeScript, Go, Agno, Milvus, BAML, AWS, Pulumi

Projects

Mobile, on-device AI

PennyWise AI

ProblemExpense trackers want your bank data in someone else's cloud.

BuiltAn Android app that reads bank SMS and categorises spending entirely on-device.

550+GitHub stars
1,000+Play Store installs

Kotlin, MVVM, on-device parsing

Guardian service catalog showing production services with tiers, stack and docs chat
Distributed systems, infrastructure

Guardian Platform

ProblemAcross many AWS accounts, nobody knows what runs where or what depends on what.

BuiltService discovery and dependency mapping from YAML manifests, with an AI docs chatbot.

100+services mapped

Next.js, PostgreSQL, pgvector, LlamaIndex

Sirius landing page
AI product

Sirius

ProblemProject work lives in scattered notes that never become assigned tasks.

BuiltAn AI project manager that turns notes and résumés into structured, assigned work. Built solo.

Livesirius.zynth.dev

LLM agents, TypeScript

HushShare: create a self-destructing secret link
Security

HushShare

ProblemPasswords and keys get pasted into chat and email, where they live forever.

BuiltZero-knowledge secret sharing: AES-256 in the browser, links that expire and self-delete.

Livehushshare.dev

I build systems that scale and AI that holds up in production: distributed backends, agents with guardrails, and evals that gate every release.

  • Python
  • TypeScript
  • Go
  • FastAPI
  • React
  • PostgreSQL
  • Docker
  • Kotlin
Backend and distributed systems
MicroservicesEvent-driven (SNS / SQS, Kafka)Distributed LLM routingMulti-tenant isolationDead-letter replayRedis cachingFastAPINode.jsFastifyHono
AI and agents
LLM agentsAgnoCrewAIMCPRAGAgent memoryGuardrailsEvalsBAMLLangfuseClaude Code
Languages
PythonTypeScriptGoJavaScriptSQL
Data
PostgreSQLMilvuspgvectorRedisKafkaDrizzle
Cloud and observability
AWSPulumiDockerCI/CDGrafanaPrometheusLokiOpenTelemetry
Web and mobile
ReactNext.jsTanStacktRPCReact NativeKotlin

Writing

All 15 posts

Hiring for backend or AI infrastructure?

hi@sarimahmed.net
Résumé