req_4417_a3f in-flight
Diwakar Patel — Portfolio Bangalore, IN Rev. 2026.08 ● Open to senior roles

Backend & distributed systems engineer.

Founding team at RagaAI. I design workflow orchestration for production healthcare automation — running live against EHRs and insurance payers, where a dropped retry stalls a real patient's claim.

Download résumé ↓ View the architecture → GitHub
RoleSDE, Founding Team
CompanyRagaAI · since 2022
DomainUS healthcare automation
CoreTemporal · K8s · Kafka
StatusOpen to roles
10Backend services
30RPA automations
4EHR integrations
99.9%Uptime sustained
Span 01 · Gateway

Four years building the systems enterprises depend on.

I joined RagaAI as part of the founding team in 2022 — ClinicalOps for US healthcare providers, and Prism and Catalyst for customers including Philips, Boeing, and Honeywell.

My work sits where distributed systems get genuinely hard: long-running workflows that span days, third-party portals that fail without warning, and multi-tenant isolation where a leak between two healthcare providers is not an option. Temporal.io, Kubernetes, and Kafka are the tools; durability and correctness are the actual job.

I also act as tech lead across squads — driving architecture decisions, reviewing code, coordinating releases, and triaging production incidents at whatever hour they arrive.

Span 02 · Workflow

ClinicalOps

A healthcare automation platform running in production for multiple US providers — insurance verification, prior authorization, fax classification, patient intake, and trial eligibility screening, end to end. Clinics run on portals that were never designed to be automated; ClinicalOps drives them anyway, durably, with credentials that never sit on disk.

Architecture owner JavaSpring Boot Temporal.ioKubernetes KafkaPostgreSQL VaultGCP
Fig. 01 — ClinicalOps, production topology Request path traced: intake → gateway → workflow → Temporal → RPA → payer
CLIENT EDGE CORE SERVICES ORCHESTRATION EXTERNAL UI / API provider portal Gateway auth · routing · tenancy Internal API Workflow clinical processes Documents LLM / Vault Temporal.io durable retries · timers workflows span days K8s Jobs 30 RPA bots one lifecycle per run EHRs eCW · Tebra · EZDerm Payer portals commercial · Medicaid Vault short-lived credentials MULTI-TENANT isolated Postgres schema per provider — onboarding needs no deploy retries survive worker restarts credentials expire, never stored
  • Durable orchestrationTemporal.io runs clinical workflows that span days — retries, timers, and compensation survive restarts and third-party outages.
  • Multi-tenant by constructionIsolated Postgres schemas, per-tenant config, and role-based access, so onboarding a provider needs no deploy.
  • Elastic prior authMoved execution from long-lived workers to Kubernetes Jobs — each run gets its own lifecycle and throughput scales with demand.
  • Credentials that expireRPA containers pull short-lived secrets from Vault instead of storing provider logins.
  • Fax triage, automatedML classification routes inbound clinical documents straight into eClinicalWorks instead of a staffed queue.
Measured impact Bars are proportional to real durations
Trial eligibility screening ↓ 96% — 5 days to 4 hours
Before
5 days
After
4 hours
Deployment pipeline ↓ 87% — 2 hours to 15 minutes
Before
2 hours
After
15 min

150 patients screened weekly · 50+ page histories · 95% accuracy

Span 03 · Execution

The rest of what I've shipped

Prism

Data quality and profiling platform. Real-time profiling backed by Elasticsearch with automated alerting and cloud storage integration.

100K+Points / day
<1msSearch
PHILIPS · BOEING · HONEYWELL

Catalyst

Evaluation platform for AI agents — automated test workflows, hallucination detection, and cost tracking across LLM providers behind a single integration layer.

99.9%Uptime
3Providers
OPENAI · CLAUDE · GEMINI · LITELLM

JobShikari

jobshikari.com ↗ Side project · in development

Built and shipped solo. An AI job co-pilot — voice mock interviews with adaptive scoring, plus job matching, resume tailoring, an application tracker, and a Chrome extension.

mic→WebRTC→Pipecat→ Deepgram STT→LLM router→ Deepgram TTS→mic
  • Realtime voice interviews. Pipecat over WebRTC with Deepgram STT/TTS, behind a self-hosted coturn TURN server so calls survive restrictive NATs.
  • Model routing by workload. Moonshot/Kimi as the general default, Cerebras zai-glm-4.7 for job scoring and live interview turns, gemma-4-31b for the fast scoring path — chosen per task on latency and cost, all behind OpenAI-compatible endpoints.
  • Polyglot, and shipped on one box. Spring Boot 3 owns the domain; a Python FastAPI service owns the voice loop, because that is where the realtime tooling lives. Postgres with Flyway-owned schema, on a single ARM64 Azure VM behind nginx and Cloudflare.
JavaSpring Boot 3 PythonPipecat WebRTCDeepgram Next.jsPostgreSQL Azure

Tools I reach for

Set large is what I lead with. Set small is what I've shipped with.

01Languages

JavaPython

Java first. Python where
the realtime tooling lives.

02Frameworks

Spring BootFastAPIMicronautDjangoFlask

Spring Boot across
ClinicalOps and JobShikari.

03Orchestration

Temporal.ioKubernetes JobsKafkaEvent-driven

Workflows that span days
and survive restarts.

04Data

PostgreSQLElasticsearchMySQLRedisInfluxDB

Schema per tenant.
Sub-millisecond search.

05Cloud

GCPAzureAWS

GCP at RagaAI.
Azure for JobShikari.

06Platform

DockerVaultTerraformJenkinsGrafanaSonarQube

Credentials that expire.
One lifecycle per run.

Span 04 · Event history

Where I've built

Software Development Engineer — Founding Team

May 2022 — Present

RagaAI · Bangalore

  • Own backend architecture for ClinicalOps; led backend development for Catalyst and Prism.
  • Acting tech lead across squads — architecture decisions, PR review, cross-team release coordination, production incident triage on GCP.
  • Multi-provider LLM integration via LiteLLM sustaining 99.9% uptime.
  • Run enterprise client onboarding from first call to live deployment.

Software Engineer & Founder

May 2020 — May 2022

Various Startups & Freelance · Remote

  • Founded Aquamarket / Borrowme — marketplace platform on Django and PostgreSQL with integrated payments and real-time messaging.
  • Built Betto — Node.js backend with a Flutter client, Unity game integration, and KYC verification.

B.Tech, Electrical & Electronics Engineering

Aug 2016 — May 2020

Noida Institute of Engineering and Technology · Greater Noida

200 OK · connection open

Let's build something durable.

I'm open to senior backend and distributed systems roles. If you're solving hard orchestration problems, I'd like to hear about them.

diwakarpatelsatya@gmail.com LinkedIn Résumé ↓