Every system I've built has the same signature: it tells you how much to trust it. I build AI systems that know what they don't know.
PDF-Rag ships with an LLM-as-judge harness that scores its own faithfulness. ChronoLens labels every claim fact, context, hypothesis, or needs-review. PathFinder hardcodes crisis phone numbers into static HTML so they outlive a backend outage. Pinch Recovery attaches a human-readable reasoning string to every decision it makes about someone's money. Krama Core runs a compliance check that returns blockers before health data moves. DeceptionArena labels each game engine by simulation fidelity so no claim outruns what's actually implemented.
None of that is the interesting part of building a model. It's the part that decides whether anyone can use it.
I'm from Andhra Pradesh (Telugu). When my grandmother needed care, our records travelled on paper. For many underserved communities in India, the nearest hospital can be 60 kilometres away; they're who this work is for. That's why I maintain Krama Core, an open-source FHIR and compliance SDK for India's ABDM health network, published under Nirvya Labs.
I tutor Data Structures & Algorithms to 50+ students a semester at the University of Sydney. Alongside that: the Data team at St Vincent de Paul Society NSW, and a university capstone with Ocius on a real-time computer-vision pipeline for autonomous maritime craft.
If you're building AI systems, developer tooling, or health infrastructure, I'd like to talk.
Nirvya Labs ·
University of Sydney ·
St Vincent de Paul Society NSW ·
Ocius Technology ·
HRVITE Services Pvt Ltd ·
Revidd ·
Open-source healthcare integration tooling
Nirvya Labs is the org the work is published under. Krama Core is its first release: a Python SDK for FHIR R4, consent and data-exchange flows, encryption, and country-aware compliance, starting with India's ABDM network and architected so the same call site works elsewhere.
Layered on purpose: clinical domains define what care is documented, country adapters define how it's identified and protected, compliance policies define what must be checked before data moves. India's ABDM adapter is the one wired for live exchange; Australia, the UK and the US carry identifier and formulary metadata only.
Apache 2.0 licensed, published to PyPI, CI across Python 3.10–3.12 with pytest, ruff, bandit and pip-audit. Contributions welcome.
A Timeless Space
A platform for Telugu cinema, built nights and weekends: editorial content, a verified marketplace of industry professionals (DOPs, music directors, editors, lyricists), craft tools, and community.
Anti-toxicity by design: no star comparisons, no box-office tracking, film-centric organisation, AI moderation on community surfaces.
Not launched.
Master of IT & IT Management (Data Analytics & Management)
University of Sydney ·
B.Tech (Honours) in Computer Science & Engineering
GMR Institute of Technology ·
AWS Certified AI Practitioner · Amazon Web Services
McKinsey Forward Program · McKinsey & Company