Harish G

Agentic AI & Automation Engineer

I build voice agents, RAG pipelines and the Java, Python and JavaScript backends they run on.

Open to AI, backend and full-stack roles.

Selected work

Rekhon

Mobile · AIFeatured

Open source · dual-licensed

Android personal-finance app that shows its working — every figure cites the rule and version that produced it. Fully offline and encrypted on the device, with no account and no server. Rupees are integer paise end to end, and a custom lint rule fails the build if a floating-point type ever holds an amount.

  • Every figure cites the rule id and version behind it, so an answer given months ago stays reproducible
  • Exact paise arithmetic, enforced by a custom Android Lint rule that fails the build rather than a reviewer who must spot it
  • Receipt OCR and bank-SMS parsing run on the device, behind per-feature consent that is off by default and revocable
KotlinJetpack ComposeRoom / SQLCipherHiltML Kit+3

CortexCraft Voice Agent

AIFeatured

Production platform for building and running Gemini Live phone agents. It gives you a dashboard to configure them, Twilio and Plivo call handling, RAG document grounding, saved transcripts, and an external REST API so other products can drive calls.

  • Gemini Live agents over real telephony (Twilio + Plivo) and a browser test channel
  • Per-agent RAG document upload, search and prompt generation
  • External REST API, so other products (CRMs, dialers) run calls through it
PythonFastAPIGemini LiveTwilioPlivo+4

LeadCall AI

AI · Full-StackFeatured

Multi-tenant outbound-calling SaaS built on the voice-agent platform. Upload a lead list and a pitch document, and Celery workers dial through the platform API, sync transcripts back and keep each tenant’s leads and credentials apart.

  • Celery dial and callback workers that call through the voice-agent platform API
  • Pitch documents re-indexed for RAG so the script matches what the client sells
  • Excel lead import, transcript sync and call-log export on Alembic-migrated PostgreSQL
PythonFastAPICeleryAlembicPostgreSQL+3

KisanVoice AI

AI · Full-StackFeatured

WhatsApp survey platform that lets farmers answer by voice in their own language. A config-driven language registry covers 24 locales (10 Indian) over Google STT/TTS, with WhatsApp Flows for long option sets, phone-call surveys through the voice-agent platform, and a real-time admin dashboard.

  • 24 locales from a config-driven registry, 10 of them Indian, with no redeploy to add one
  • Migrated the platform off a Node backend and off Sarvam onto FastAPI + Google STT/TTS
  • 172 of the repo’s 192 commits; Jenkins CI, Docker Compose deploys
PythonFastAPIReactMongoDBWhatsApp Flows+5

FinGraph AI

AI · Full-StackFeatured

Banking assistant over a Neo4j relationship graph. pgvector semantic search finds the relevant customers, loans and chats, Neo4j enriches every match with its relationships, and Groq writes the answer. JWT + bcrypt authentication guards the admin and audit APIs.

  • Hybrid retrieval: pgvector similarity search enriched with Neo4j relationships
  • MiniLM embeddings from Hugging Face, answers generated with Groq
  • JWT + bcrypt authentication, with ADMIN-only admin and audit routes
Node.jsExpressReactNeo4jSupabase pgvector+3

About me

I have been a Software Developer at CortexCraft.ai since January 2026, after several years of freelance web and app work. I write backends in Spring Boot, Express.js and FastAPI, and connect them to LLMs, speech-to-text, text-to-speech and tool-calling agents.

Core toolkit

LLM IntegrationRAGContext ManagementAgents & Tool CallingOutput GuardrailsWorkflow AutomationSTT / TTS SystemsGemini LiveLightGBM / ProphetSpring BootFastAPIReactNeo4jDocker