🌾 Kheti Sahayak — AI-Powered AgriTech Platform
Empowering 1M+ Indian farmers with AI diagnostics, hyperlocal weather, digital marketplace, and expert consultations.
What it is: Production SaaS platform for Indian smallholder farmers — crop disease detection via vision AI, real-time mandi prices, expert consultations, and a full digital marketplace.
Architecture highlights:
- 28 Spring Boot microservices (100% parity migration from Node.js monolith) — auth, marketplace, crop, weather, diagnosis, notifications, payments, and more
- 162 AI agents defined as Claude Code subagents — engineering, QA, product, DevOps, security, ML, marketing, executive roles — orchestrated via CrewAI crews and a Virtual Software Company pattern
- Claude Code deeply integrated — 162
.claude/agents/definitions with YAML frontmatter, hooks (SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, Stop), GPS MCP server for symbol graph indexing, Render MCP for deployments - Multi-model AI stack:
claude-sonnet-4-6(default),claude-opus-4-8(high-reasoning CFO/CTO roles),claude-haiku-4-5(lightweight ops), Cerebras (inference), Groq (audio transcription), Gemini - LLaVA vision model — local CPU inference for crop disease detection across 38 PlantVillage classes, with 2000+ lines of India-specific treatment database (Hindi/English/Telugu/Marathi)
- Multi-tenant SaaS — PostgreSQL schema isolation per tenant, X-Tenant-Id header enforcement, row-level security
- Observability: ELK Stack, Prometheus + Grafana, Spring Sleuth + Zipkin, Sentry
- Event streaming: Kafka 3.7, Redis 7, PostgreSQL 14 + pgvector
Agent-driven SDLC — 110+ workflows covering the full engineering lifecycle:
| Phase | What's automated |
|---|---|
| Planning | 20 parity audit plans (auth → marketplace → weather → finance → profile), sprint backlog, PRD templates, architecture diagrams, UI specs |
| Development | 9 agent-chain workflow types (WF-1 new feature → WF-9 backend bug fix) — each dispatches multi-role chains: Product Manager → Architect → Backend Dev → Mobile Dev → QA → Security Engineer |
| Testing | Full-app emulator QA sweep: 7 sequential runtime agents + 16 parallel static auditors → P0/P1/P2/P3 defect triage → parallel fan-out fix dispatch per defect group |
| Deployment | Module-aware deploy (WF-10/11/12): detect changed modules → Docker/Render/Vercel/Play Store → health verify → auto-rollback on failure |
| Monitoring | 24/7 L1→L2→L3 agent escalation hierarchy: health scan → triage → remediate → investigate → stabilize → incident command → post-mortem |
| CI/CD | 30 GitHub Actions: multi-stack CI (Node.js, Spring Boot, Flutter, Kotlin), AI-powered PR code review, CodeQL SAST, Firebase Test Lab on real devices, smoke tests on every production push |
| Self-improvement | Nightly agent improvement loop: harvest GPS corrections → triage by agent → mutate prompts below 7.5 threshold → commit; real-time auto-learn stores feedback instantly; GEPA-style eval loop scores agents across 5 dimensions |
Workflow runner: single ./workflow-runner.sh WF-N TASK-ID dispatches full agent chain, logs every step to shared GitHub issue as paper trail.
Stack: Java 17 / Spring Boot 3.4 · Node.js 18 / Express 5 · Flutter 3 · React 19 + Vite · Python / FastAPI · PostgreSQL · Redis · Kafka · Docker / Kubernetes · Render · Vercel
🤖 Job Search AI SaaS Application — Agentic Job Discovery & Auto-Apply
End-to-end agentic pipeline: discover → ATS-score → auto-apply across Naukri, LinkedIn, and Indeed.
What it is: Multi-tenant job-search platform that started as a single-admin Notion CLI tool and grew into a full SaaS layer with magic-link auth, per-user quota tiers, an LLM tailor pipeline, a chat agent, and a React frontend — all in one Express process.
Agentic highlights:
- Claude Code subagent fleet — 15 specialized agents per portal (session-manager, job-discoverer, job-scorer, cover-letter-writer, apply-bot, pipeline-orchestrator) across Naukri / LinkedIn / Indeed
- Auto-discover → ATS-score → auto-apply pipeline — Playwright-driven headless apply with stealth mode, session persistence, and smart form-fill via ApplyAgent
- 11-step LLM tailor pipeline — JD Analysis → Keywords → Company Research → Resume Match → Skill Gap → Rewrite Bullets → Cover Letter → Interview Qs → Recruiter Outreach → Critic → Revision → Finalize; hybrid ATS scoring with automatic revision loop
- Chat agent — router → specialist delegation (discovery / scoring / tailor / apply / resumes / schedule) over Anthropic SDK, tool-use loops, capped at depth 4
- Self-improving apply agent — learns from outcomes; persists qa_cache, selector_drift, portal_overrides, outcomes.jsonl
- SSE streaming for long-running operations; SQLite (better-sqlite3, WAL mode) for both app and pipeline state
Stack: Node.js 18 / Express · React 18 + Vite + TypeScript · Python / Playwright + playwright-stealth · SQLite · Anthropic SDK · JobSpy · Resend · Razorpay
- 🔭 Kheti Sahayak — scaling to 1M+ farmers, completing microservices migration, adding pgvector RAG
- 🤖 Agentic systems — Claude Code multi-agent orchestration, CrewAI crews, self-improving bots
- 🌱 Spring Boot 3.4 + Kafka microservices architecture at production scale
Java · Spring Boot · Node.js · Microservices · Claude Code & Agentic AI · Playwright automation · Multi-tenant SaaS · PostgreSQL · Kafka · Docker/Kubernetes
Email: p.ponali@yahoo.com
Blog: prakashponali.wordpress.com
