Issue · 01 · MMXXVI

Hyderabad · IN

Open to applied scientist roles

Engineer,
mentor,
builder&writer.

A note

I build AI systems that work outside the lab — from edge-deployed computer vision on poultry farms to enterprise NLP pipelines at S&P Global. I mentor the next wave of agritech engineers at LiveStockify, and I write about what I learn along the way.

explore my work read the blog

§ 01 — Who I am

An engineer with roots
in the real world.

I care about systems that actually run — on constrained hardware, against messy data, in places far from the cloud. The interesting problems are the ones the demos never show.

01Engineer

S&P Global

Senior AI/ML Engineer on Enterprise Technology. AWS Step Functions, OpenSearch, Azure OpenAI — five years of shipping systems that scale to the enterprise.

02Mentor

LiveStockify

Co-founder and mentor. Guide a team of engineers building AI-powered poultry behavioral monitoring at the edge — agritech meets deep learning, live on real farms.

03Builder

Research & beyond

Targeting Applied Scientist roles at Amazon Science, Microsoft MAI, and Sarvam AI. Publishing VLM work on HuggingFace and arXiv workshop tracks.

— Tools of the trade

Python YOLOv8 FastAPI AWS Lambda OpenSearch Azure OpenAI Step Functions Raspberry Pi 5 Ollama Moondream 2B LoRA Docker PyTorch Tailscale

§ 02 — Selected work

Things I've shipped.

→ 01

Agritech · Edge AI · Startup

LiveStockify — poultry behavior monitor

End-to-end YOLOv8 system detecting Eating, Drinking, Sitting, and Standing across 923 broiler birds on 8 RTSP cameras. Live on Raspberry Pi 5 via Tailscale VPN. FastAPI backend with API key auth. Mentoring three engineers building on top of this foundation.

YOLOv8 Raspberry Pi 5 FastAPI RTSP Tailscale mAP50 0.454

↗

→ 02

Enterprise · AWS · NLP

Contract Terms Tracker — S&P Global

AWS Step Functions pipeline integrating Lambda, OpenSearch vector search, and Azure OpenAI. Reduced search latency from 9s to 5.5s via concurrent LLM summarization, embedding caching, smart k-amplification, and ef_search tuning. Comprehensive S3 failure tagging across all pipeline lambdas.

AWS Lambda OpenSearch Azure OpenAI Step Functions FastAPI

↗

→ 03

Research · VLM · Edge

LiveStockify Vision — VLM phase

Docker Compose stack with Ollama and FastAPI running Moondream 2B as a broiler farm inspector. ~2.4s per image on Apple Silicon. LoRA fine-tuning underway, with an arXiv workshop paper (CV4Animals / Agriculture-Vision) and HuggingFace release in progress.

Moondream 2B Ollama LoRA Docker VLM arXiv

↗

§ 03 — Writing

Thinking out loud.

— Coming soon edge ai

Deploying YOLOv8 on Raspberry Pi 5 — what nobody tells you

RTSP streams, Tailscale VPN, and 923 live chickens. What I learned about real edge deployment.

soon

— Coming soon engineering

Cutting OpenSearch latency in half: a field guide

How concurrent LLM calls, embedding caches, and ef_search tuning took us from 9s to 5.5s.

soon

— Coming soon mentorship

Mentoring junior engineers in a startup — what actually works

Lessons from building and guiding a small agritech CV team from zero to live farm deployment.

soon

— Coming soon career

The 2026 AI/ML job market in Hyderabad — honest research

Staying vs. leaving, research visibility, and targeting Applied Scientist roles at the big names.

soon

§ 04 — Personal

Life beyond the terminal.

@nannorio · Instagram

Telugu traditions
meet modern science.

A Telugu father documenting his daughter's first years — evidence-based, culturally rooted, and honest. Where ancient wisdom and peer-reviewed research share the same feed.

→ 01Six-month milestones — what the research actually says.

→ 02Introducing solids — Ayurveda meets the WHO.

→ 03Baby in unexpected context — the viral Reel formula.

→ 04First home, first daughter — same chapter of life.

§ 05 — Get in touch