Loading…
Hire me

DineshKumarG

Software Engineer · Bangalore, India

I build multi-agent AI systems for healthcare.

01 — About

Agents that do real work, in a field where it matters.

Software Engineer at HealthFoundry.ai building event-driven, multi-agent AI systems for healthcare. I ship distributed backend services in Python and TypeScript — LangGraph orchestration, Pub/Sub messaging, real-time delivery and human-in-the-loop routing — on GCP and AWS, and I own the notification, observability and platform SDK layers used across the company's products.

8services on my observability stack
2roles at HealthFoundry.ai
8.94CGPA · BMS College of Engineering

02 — Experience

HealthFoundry.ai

Software Engineer

Jul 2026 — Present

HealthFoundry.ai (SynapseNet)

  • Designed a LangGraph orchestrator that takes work over REST and Pub/Sub and runs declarative workflows stored in Firestore — resolving steps per request type, dispatching to the right agent and routing results downstream.
  • Built document- and speech-intelligence agents on Gemini 3.7 Flash — leads, tickets and follow-ups from call recordings; clinical insights from discharge summaries; medications and conditions from prescriptions.
  • Closed the loop from agent to CRM: extracted entities are written back and a notification ships with a deep link to the new lead.
  • Own the observability stack (OpenTelemetry, Tempo, Loki, Prometheus, Grafana, MLflow) across 8 FastAPI and Fastify services, on AWS EC2 with GitHub Actions CI/CD.
  • Authored the shared Platform SDK — auth, profile, permission and domain lookups behind one interface, cutting integration to a few lines.

Data Engineer Intern

Sep 2025 — Jun 2026

HealthFoundry.ai (SynapseNet)

  • Built the Human Interaction Layer — an event-driven service consuming CloudEvents from Pub/Sub, resolving recipients against Firestore rules and pushing real-time notifications over SSE.
  • Built the clinical document ingestion pipeline — LangGraph, PHI/PII de-identification with Gemini, Medplum FHIR mapping and SNOMED CT coding via HL7 terminology servers.
  • Developed the Chrome extension assistant with one-click actions: form filling, opening linked records, authorizing or dismissing notifications.
  • Managed Firebase hosting, Firestore and Cloud SQL schemas and GCS bucket operations.

03 — Skills

The toolkit.

Languages

PythonTypeScriptSQLC++

AI / LLM

LangGraphLangChainCrewAILangSmithMLflowMCPRAGHugging FacePyTorchTensorFlowGeminiGroq

Architecture

Event-drivenMicroservicesPub/SubCloudEventsSSEHuman-in-the-loopMulti-agent orchestrationSDK design

Backend & Frontend

FastAPIFastifyNode.jsExpressReactREST

Cloud & Ops

GCPFirebaseFirestoreCloud SQLAWS EC2DockerKubernetesGitHub Actions

Observability

OpenTelemetryGrafanaTempoLokiPrometheus

Data

MongoDBMySQLChromaDBFAISS

Healthcare

FHIR (Medplum)HL7SNOMED CTPHI/PII de-identification

04 — Projects

Things I've built.

01

HealthFoundry · production

Workflow Orchestrator

LangGraph engine executing declarative, Firestore-stored workflows over REST and Pub/Sub, dispatching each step to the right AI agent.

LangGraphPub/SubFirestore
02

HealthFoundry · production

Human Interaction Layer

Real-time routing service: CloudEvents in, permission-aware notifications out over Server-Sent Events.

CloudEventsSSEHITL
03

HealthFoundry · production

Clinical Document Pipeline

Discharge summaries to FHIR resources — classification, de-identification and SNOMED CT coding.

GeminiFHIRSNOMED CT
04

Personal · Aug–Sep 2025

LangGraph AI Chatbot

Stateful multi-threaded chatbot with persistent memory, Groq LLM tools (web search, stocks, math), streaming and SQLite checkpointing.

LangGraphLangSmithStreamlit
05

Personal · May–Jun 2025

YouTube Transcript Q&A RAG

RAG over video transcripts — FastAPI ingestion, Cohere embeddings, FAISS, ChatGroq. Dockerized on AWS EC2.

RAGFAISSDocker

05 — Education

B.E. Computer Science & Engineering

BMS College of Engineering, Bangalore · 2022 — 2026

CGPA 8.94

Certifications

  • Python for Data Science — NPTEL
  • Machine Learning Professional — RapidMiner

06 — Contact

Let's work together.

Open to roles and conversations in AI engineering, agent platforms and healthcare tech.

© 2026 Dinesh Kumar G