Hello, I'm Dhvani Raval. An —

Dhvani Raval, AI engineer who ships, evaluates, retrieves, automates and monitors.

AI and ML engineer who takes models from research to revenue.

Ten-plus years across data, machine learning and generative AI, from SQL pipelines at enterprise scale to autonomous LangGraph agents running on Cloud Run, now shipping agents, RAG and classifiers across 1,000+ dealership platforms at Sincro.

I believe a model only counts once it's evaluated, monitored and paying for itself, and I'm here to keep shipping ones that do.

Shipped with

Selected work

Systems in production

LangGraph · Gemini Pro · Cloud Run

Autonomous BI Dashboard Agent

Business questions in, verified SQL and answers out, over multi-terabyte warehouses.

NL → SQL on TB-scale schemas

LLM-as-a-judge · Evals

Conversational Outcome Classifier

Cheap rules triage first; the LLM only sees what's ambiguous.

~20% lower API cost

RAG · GA4 · GTM

Tracking-Spec RAG Assistant

Structure-aware chunking and hybrid search over multi-tab specs.

65% faster ticket turnaround

PyTorch · DistilBERT · ONNX

Real-time Intent Engine

Fine-tuned, quantised and exported to ONNX with automated retraining.

10 intents · sub-50ms

XGBoost · Evidently AI · Vertex AI

Vehicle Attention Scoring

Purchase-readiness scoring, monitored for drift, retrained automatically.

$1M+ estimated annual efficiency

Pipelines · Summarisation

Dealer Intelligence Summarizer

Condenses multi-source reports into briefs a human can act on.

Reports → briefs, automatically

02 · How I think

Production, not demos

The model is the easy part. Evaluation, cost and monitoring decide whether it survives contact with real users.

Principle 01

Evaluate before you ship

LLM-as-a-judge, agreement benchmarks and golden sets, so "it seems better" becomes a number.

Principle 02

Cost is a model metric

Hybrid routing, rule-based fallbacks and caching. The cheapest correct answer wins.

Principle 03

Watch it after launch

Drift detection, monitoring and auto-retraining, because the data never stays stationary.

Automotive AIProperty managementAdvertising technology

Education

PG Diploma in Data Science

Symbiosis University · machine learning, deep neural networks, big data and applied NLP.

Education

B.E., Electronics & Communication

VGEC, Gujarat Technological University.

03 · Toolkit

The toolkit

Languages

PythonSQLR

LLMs and providers

GeminiClaudeHugging FaceOpenAIOllamaMistral

Agents and RAG

LangGraphLangChainMCPPineconeChromaQdrantLlamaIndex

ML and deep learning

PyTorchTensorFlowscikit-learnXGBoostONNXNumPypandasKerasJupyter

MLOps and data

Evidently AIMLflowDatabricksSparkBigQueryWeights & BiasesAirflowDVCKafkaRayStreamlitGradio

Cloud and infrastructure

GCP · Vertex AI · Cloud RunAzure MLDockerKubernetesFastAPIGitHub ActionsGitTerraformRedisNVIDIA CUDAGrafanaPrometheus

Yellow = core stack

Measured outcomes

By the numbers

$1.0M+Annual opex reduction

92%LLM-judge agreement

20%Inference cost drop

15+Models in production

Experience

Where I've shipped

2021 — now

Software Engineer, AI & ML

Architecting, fine-tuning and deploying generative AI systems, agentic workflows and predictive models across 1,000+ dealership platforms on GCP: an autonomous BI agent, an attention-scoring engine worth $1M+ a year, and an LLM-judge classifier at 92% agreement.

Sincro, a DealerOn companyPune, India

2018 — 2021

Senior Software Analyst

Where I learned to respect clean data: SQL transformation pipelines, automated validation that cut manual audits by over 40%, and indexing patterns for multi-tenant scale.

Yardi SystemsPune, India

2016 — 2018

Senior Specialist, Data & Analytics

Statistical lead-scoring models in Python and R, audience segmentation, and ETL from marketing APIs and CRM endpoints into ROI dashboards.

Selling SimplifiedPune, India

2014 — 2016

Software Research Analyst

Data-mining roots: Python crawlers harvesting market intelligence, hypothesis tests holding 98%+ dataset integrity, and ETL cleaning that cut ingestion latency by 35%.

AnosalesPune, India