Aditya Katheeth

AI & Data Engineer · LLM Systems · ML Infrastructure · Applied AI Research

Production data systems by day. Researching how the systems we ship affect the people they decide about.

8+ years shipping production data infra at scale
M.S. in AI · transitioning to AI engineering
Fairness audits as a craft, not just a topic

See production work →Get in touch

Now

In production right now

Updated quarterly. The honest version.

▸Architecting agentic AI systems on Snowflake Cortex
▸Reading: Crawford's Atlas of AI; Barocas/Hardt/Narayanan
▸Building: an open-source fairness audit playground
▸Applying: doctoral programs in AI ethics

Featured Work

Production AI in the wild

Three systems, live and measured.

AI Engineering

Most Recent

Snowflake Cortex Multi-Agent Platform

Multi-agent LLM system over enterprise data. Semantic views deployed to enable natural-language querying across three business units, with full auditability and access controls. Outputs surfaced at VP level.

OutcomeVP-level adoption across 3 business units

Snowflake CortexMulti-AgentLLMSemantic Layer

AI Engineering

Most Recent

Real-Time NLP Sentiment + Market Intelligence

Production NLP pipeline combining customer sentiment analysis with competitive market intelligence. End-to-end: ingestion, transformer-based sentiment scoring, entity linking, and aggregated reporting.

OutcomeReporting turnaround cut from 2 weeks to 2 days (85% faster)

NLPStreamingPythonHuggingFace

MLOps

Most Recent

ML-Based Anomaly Detection on Snowflake

ML system continuously monitoring query execution patterns across thousands of daily runs. Surfaced model degradation, data integrity failures, and performance regressions before they reached downstream AI systems.

OutcomeCaught 12+ production-impacting regressions before user impact

MLOpsAnomaly DetectionTelemetryDrift Detection

See all production work →

Tools

Open source — built so engineers can use it

Fairness Audit Playground

Drop in a CSV. Get a fairness audit in under a minute.

Engineering hygiene for ML teams shipping to production — before researchers tell them to. Built so disparate-impact checks happen by default, not by accident.

Try it →

Writing

Notes from the production trenches

Engineering, ML, and the systems we ship.

Coming soon — drafts in flight.

Browse all writing →