Headline

AI first Engineering Leader · LLMs · RAG · RLHF · Agentic AI · ASR/TTS · Multimodal · 2M+ Farmers · 10+ Languages · FarmerChat @ Digital Green · Building AI that works where the internet barely does

About

AI built for 2 million farmers — in their language, on their device, in their world where the internet barely works. That's FarmerChat. I lead the engineering behind it. Getting a model to work in a demo is easy. Getting it to work for a farmer in Odisha — in odia, on a ₹5,000 Android, with 2G connectivity, with a question that has real consequences — that's a different problem entirely. At Digital Green, I've spent 3 years solving exactly that. The AI stack runs deep: - LLMs & Generation: Production LLM pipelines for high-stakes agricultural advisory — prompt engineering, context management, hallucination mitigation, response grounding - Retrieval & RAG: Multilingual RAG with retrieval reranking across low-resource languages — because a hallucination about a pesticide dose isn't a UX problem, it's a safety problem - Fine-tuning: RLHF on golden QA datasets curated by domain agronomists and MEL teams — model behaviour shaped by field expertise - Agentic AI: Multi-step reasoning workflows integrating weather APIs, market prices, Plantix, and crop calendars — the system resolves complex farmer queries autonomously across tools - Speech: Multi-vendor ASR/TTS — Google Chirp, Bhashini, Addis AI and others — language-optimised per vernacular across 10+ languages, because no single vendor wins every language - Multimodal: Text, voice, and image interfaces — literacy and connectivity aren't prerequisites - Analytics & Engagement: PostgreSQL database, Bigquery datawarehouse, MoEngage and Plotline for in-app engagement and personalisation · Adjust for marketing attribution and growth tracking — each tool puroseful at scale. - Infrastructure & Platform: Microservices architecture on FastAPI, distributed systems deployed on AWS EKS — built for scale, resilience, and independent service deployability 3 million users across two products. Before FarmerChat — BESCOM Mithra, 1 million citizens, Chief Minister of Karnataka recognition. I hire for curiosity over credentials and build AI literacy across the whole team — no black boxes, no single points of failure. Cross-functional alignment with product, marketing, and MEL is what keeps the impact at scale — not just the code. 14 years across the full stack. Enough depth to know where things break. Enough leadership to build teams that fix them. AI & Data: LLMs · RAG · RLHF · Agentic AI · ASR/TTS · Multimodal · MoEngage · Firebase Engineering: Python · Django · PostgreSQL · AWS · Android PG in AI & ML. Oracle Certified Java SE Programmer. If you're building AI that needs to work beyond the demo — let's talk.

Location

Bengaluru

Work experience

Engineering Manager at Digital Green, Apr 2024 - Present · 2 yrs 2 mos

Assistant Engineering Manager at Digital Green, Apr 2023 - Apr 2024 · 1 yr 1 mo, Bengaluru, Karnataka, India

Technology Lead at Digital Green, Apr 2022 - Jul 2023 · 1 yr 4 mos, Bengaluru, Karnataka, India

Technical Lead Manager at BESCOM, Feb 2015 - Apr 2022 · 7 yrs 3 mos, Bengaluru, Karnataka, India

Software Engineer at ROBERT BOSCH GmbH, Jul 2012 - Feb 2015 · 2 yrs 8 mos, Bangalore

Education

Post Graduation Diploma, Artificial Intelligence and Machine Learning, NIT warangal

Bachelor of Engineering - BE, Computer Science, Bangalore institute of technology