I build AI systems auditors can trust — schema-constrained extraction, citation-grade RAG, deterministic-first pipelines, and on-prem AI when compliance demands it.
12+ years across data pipelines, distributed systems, and applied AI. Lead AI engineer on document intelligence and LLM workflows for financial audits at Radar360. Previously scaled commerce data systems at Hearst Magazines.
“A flagged ‘needs review’ beats a confident wrong answer.”
The principle behind everything I build.
Impact at Scale
~10x
AI Q&A cost cut via query routing
99.4%
offline mapping accuracy, AI off
62→16GB
31B quantized, full 256K context
2 wks
offline AI product, spec → installer
15M+
products processed daily
10M
pages crawled / day
$200K
saved via AI matching
60+
services migrated to Kubernetes
What I've Built
View all →Offline AI Desktop Product
Air-gapped desktop app — deterministic-first matching with an embedded local LLM, 99.4% accurate with AI off
Fine-tuning Gemma 4 31B
On-prem document analysis on 100% synthetic data — QLoRA → GPTQ → vLLM for ~$63
Document Intelligence Platform
LLM-powered document classification, extraction & Q&A with AWS Bedrock
Multi-Tenant Analytics Platform
Database-per-tenant backend with async workflows and multi-DB querying
Dynamic Report Generator
Template-driven Excel generation service powering 40+ report types
E-Commerce Data Pipeline
15M+ SKUs/day ingested from hundreds of retailers via FTP, SFTP & APIs
Distributed Web Crawler
400-node crawler processing 10M pages/day across 300+ websites
Where I've Worked
|
Have a question? Say hello.