Vimal Nakrani — Data & AI

HOUSTON, TX — DATA & AI

Vimal Nakrani

Enterprise-grade by day. Open-source by night.

Data & AI Manager at EY. I design lakehouse platforms on Databricks and Azure, and ship applied AI — RAG, agents, MCP — that turns complex data into business outcomes.

After hours I publish open models and tooling for auditable AI: benchmarks, evals, quantization ladders. Everything measured, nothing vibes-based.

See the work ↓ Get in touch
7+
years in Data & AI
13
models on Hugging Face
1.6K+
downloads on Hugging Face
11
open-source repos
0
PyPI downloads (3 libraries)

01 / ABOUT

One brain, two modes.

The same instinct runs through both: systems you can trust because you can inspect them.

BY DAY

The enterprise builder

Leading Data & AI delivery at EY — lakehouse architecture, governed pipelines, and applied-AI systems that have to work for real businesses, at real scale, under real scrutiny.

Databricks & Azure lakehouse platforms
RAG, agents & MCP integrations in production
From C-suite outcomes to engineering reality

BY NIGHT

The open-source builder

Building the tooling that makes AI auditable — open audit-domain models, tamper-evident RAG trails, eval harnesses, and quantization ladders where every claim ships with a measurement.

Auditable AI: AuditLM, AssuranceBench, AuditWeave
Local inference & quantization (MLX / GGUF)
LLM evals, benchmarks & fine-tuning

// the overlap: I work at an assurance firm — and I build tools
// that make AI systems auditable. Same instinct, different stacks.

02 / EXPERIENCE

Where the day job happens.

Seven-plus years from data pipelines to AI platforms.

Data & AI Manager

CURRENT

EY · Assurance Services

Aug 2024 – Present · Houston, TX

Leading data & AI for EY Assurance — architecting the audit data platform and shipping production LLM systems used by 1,000+ engagement teams.

Architected an enterprise audit data platform — Azure Data Factory, ADLS, Databricks, Delta Lake, Unity Catalog — with end-to-end lineage from ERP sources to Power BI, across 8 workloads supporting 1,000+ engagement teams.
Cut end-to-end pipeline latency 30–40% through distributed PySpark and Delta Lake optimization — joins, partitioning, caching, workload design.
Shipped production LLM capabilities across audit workflows — RAG, agents, fine-tuned models, vector retrieval, evals, guardrails — reducing manual effort ~60% across 5 workflows.
Built a read-only natural-language query assistant — a LangGraph ReAct agent on Azure OpenAI over PostgreSQL and Databricks — with a three-layer SQL safety framework and PCAOB AS 2401 journal-entry anomaly detection.
Led a 6-engineer global team delivering AI initiatives end to end — architecture, delivery, stakeholder alignment — while setting CI/CD and engineering standards in Azure DevOps.
Databricks Azure Unity Catalog RAG Agents LangGraph LLM Evals

Senior — Analytics & Data Engineering

EY · Assurance Services · Oct 2021 – Jul 2024

Re-engineered Alteryx workflows into optimized Spark and Databricks pipelines — faster processing, lower infrastructure cost.
Processed multi-terabyte client datasets — 5+ TB through a single pipeline — and cut a high-risk healthcare revenue analytics process by ~40%.
Managed 10 concurrent projects for a Fortune-listed client across 10 countries; trained 15 team members and established the team's SDLC in Azure DevOps.

Data & Analytics Specialist

EY · Assurance Services · Nov 2019 – Sep 2021

Automated weekly financial reporting with Alteryx (~70% less manual effort) and complex financial calculations with Python.
Designed SQL ETL pipelines from client ERP systems; built Power BI and Tableau dashboards surfacing trends, anomalies, and KPIs.
Developed a PySpark early-risk-identification process that surfaced potential issues earlier in the audit lifecycle.

EDUCATION

MS, Computer Science

University of Houston–Clear Lake · 2019

EDUCATION

BE, Information Technology

Mumbai University · 2017

03 / OPEN SOURCE

Built in the open. Measured in public.

Open models, benchmarks, and tooling for AI you can actually trust — every release benchmarked before it ships.

THE AUDITABLE-AI STACK

auditlm

AuditLM

Open, locally-runnable language model and benchmark for US GAAP & PCAOB-based external audit.

Python · Fine-tuning · Local LLM

assurancebench

AssuranceBench

Benchmark for evaluating language models on external-audit work — with a safety and guardrail suite.

Python · Benchmarks · Guardrails

auditweave

AuditWeave

Tamper-evident audit trail for RAG pipelines — trace any AI conclusion back to its evidence.

Python · RAG · Provenance

PyPI downloads

MORE TOOLS

dframe-trace

Find where your pandas/polars pipeline silently broke — nulls, dropped rows, dtype drift. No rules to write.

PyPI downloads

RegRadar

AI-powered search and alerts for the US Federal Register.

unlimited-ocr-eval-harness

Synthetic OCR corpus + CER/WER harness for measuring exactly what quantization costs (MLX + GGUF).

stillwarm

KV-cache persistence for local LLMs — stop re-paying prompt processing after every restart. Live demo on Spaces.

PyPI downloads

Download counts via pepy.tech (official PyPI BigQuery data), refreshed daily.

Model downloads on Hugging Face

13 models · 2 datasets · 4 spaces · refreshes daily

unlimited-ocr-gguf
968
unlimited-ocr-4bit-mixed-mlx
146
unlimited-ocr-6bit-mlx
129
unlimited-ocr-bf16-mlx
129
unlimited-ocr-8bit-mlx
87

+ 8 more variants, incl. the HY-Embodied-0.5 MLX ladder (4/5/6/8-bit & bf16) — every quant published with a measured eval ladder, not vibes.

DATASETS

View all on Hugging Face →

04 / SKILLS

The working stack.

DATA & PLATFORMS

{{ c }}

GENERATIVE AI & LLMs

{{ c }}

TOOLS, APPS & RESEARCH

{{ c }}

05 / PUBLICATIONS

Research on record.

PAPER

AuditWeave: A Tamper-Evident, Auditor-Navigable Evidence Layer for AI-Assisted and Data-Transformation Workflows

arXiv:2607.09682 — the research behind the open-source AuditWeave project; provenance you can hand to an auditor.

REVIEW

A Review: Internet of Things (IoT) Based Smart Home Automation

Survey of IoT architectures and protocols for smart home automation.

All publications on Google Scholar →

06 / CERTIFICATIONS

Certified, not just self-taught.

DATABRICKS

Databricks Certified Data Engineer Professional

Issued May 2026 · Expires May 2028

Credential ID 183360013 · Verify →

MICROSOFT

Microsoft Certified: Azure AI Fundamentals

Issued Mar 2024 · No expiry

Credential ID 34D2742438262E4E · Verify →

07 / WRITING

Notes from real use.

Overlooked fundamentals and new AI tools I've personally tested — everything from real use, not announcements.

DEV.TO

I Tested Quantized Unlimited-OCR on Mac. 4-Bit Was Not the Sweet Spot.

CER/WER ladders across MLX quants — and why mixed 4-bit beat straight 4-bit.

Jul 2026 · 7 min read

DEV.TO

Why did my DataFrame lose rows? Debugging silent pandas pipeline failures

Tracing silent row loss, nulls, and dtype drift through pandas pipelines.

Jun 2026 · 5 min read

HUGGING FACE

Your local LLM forgets its cache after every restart. I measured how much time that wastes.

The real cost of cold KV-caches in local workflows — and how stillwarm fixes it.

Jul 2026 · Article

07 / CONTACT

Let's build something that holds up.

Always up for genuinely hard data problems. If you need AI that can survive an audit — literally — say hi.

Connect on LinkedIn
LinkedIn GitHub Hugging Face dev.to Google Scholar
© 2026 Vimal Nakrani · Houston, TX enterprise by day · open source by night