Available for ML / AI Roles•Haryana / Delhi-NCR, India

MACHINE LEARNING • APPLIED AI • DATA SYSTEMS

MACHINE LEARNING ENGINEER

Gourav
Gulia

Machine Learning Engineer building production-grade ML, GenAI, and data systems.

Machine LearningSimilarity SearchGenerative AIModel Serving

Work Experience Featured Projects Resume (PDF)

Professional Highlights

EY • PUBLIC SECTOR

Enterprise AIEY (Ernst & Young)

Built production face verification and forensic AI systems for national government clients.

Similarity Matching1M+ Records

Fast similarity search across high-volume databases to identify duplicate identities.

Production PipelinesFast APIs

High-performance backend services, automated workflows, and version-controlled data pipelines.

Reliable • Production ReadyDelhi-NCR, India

ROLE

Sr. Analyst

Enterprise AI & Machine Learning at EY

SCALE

1M+ Records

Fast similarity matching across large databases

SYSTEMS

Production APIs

Low-latency services and automated workflows

PRACTICE

Tested & Versioned

End-to-end reproducible data pipelines

EXPERIENCE

Work Experience

Professional history building enterprise ML systems and data pipelines.

Professional Summary

Machine Learning Engineer with hands-on experience building and deploying production ML systems at Ernst & Young (EY). Specializing in MLOps pipelines, high-dimensional vector search, and multi-agent GenAI architectures. Backed by a strong mathematical and statistical foundation (M.Sc. Data Science) with a focus on low-latency inference, reproducible pipelines, and clean software design.

Companies

Education

Master of Science (M.Sc.) in Data Science

Chandigarh University

Jul 2022 – May 2024 • Chandigarh, India

Bachelor of Science (B.Sc.) in Applied Science

Delhi University

Jul 2019 – May 2022 • Delhi, India

EY (Ernst & Young)Jun 2025 – Jun 2026

Sr. Analyst — AI & Machine Learning Systems

Enterprise forensic AI, biometric verification pipelines, and automated MLOps for large-scale public sector clients (SSC, HSSC, NHA).

Key Accomplishments:

Developed and deployed production face verification, image quality assessment, and identity fraud detection pipelines for national government clients.
Designed an ensemble model evaluation framework with calibrated threshold tuning, reducing false positive detections by nearly 11%.
Built high-performance similarity search across 1M+ candidate records and text matching systems, improving duplicate identity detection by ~25%.
Modernized legacy ML systems into modular backend services with automated workflows; optimized model inference to reduce latency by ~35% and memory usage by 22%.

Search Scale

1M+ Profiles

Orchestration

Automated Pipelines

Serving Latency

~35% Reduction

Deployment

Containerized Services

STACK:PythonFastAPIApache AirflowMLflowDeep LearningComputer VisionOpenCVVectorDBDockerScikit-LearnPandas

PROJECTS

Featured Projects

Production systems, multi-agent architectures, and developer tooling.

Multi-Agent AI•2026

OPEN SOURCE

Analytica: Multi-Agent AI Data Analysis System

Autonomous data analysis with stateful multi-agent orchestration and Python execution

Autonomous multi-agent system combining LLM reasoning with sandboxed, deterministic Python execution for verified tabular data analysis and visualization.

Built a modular agent architecture using LangGraph for stateful query planning and isolated code execution.

Implemented structured Pydantic validation and error recovery loops to ensure reliable, code-grounded results.

GitHub Repository

Orchestration

Stateful LangGraph

Execution

Sandboxed Python

Architecture

Multi-Agent Personas

Validation

Pydantic Schemas

TECH:PythonLangGraphLangChainGroq LPUPydantic v2StreamlitDocker

MLOps & Systems•Jun 2026 – Aug 2026

DEPLOYED

BurnoutAI: End-to-End ML Risk Prediction Engine

Configurable ML pipeline with versioned datasets, experiment tracking, and Docker serving

End-to-end machine learning system translating student behavioral and academic data into calibrated burnout risk scores, with complete artifact versioning and containerized serving.

Built an artifact-driven pipeline using DVC and MLflow for deterministic version control from raw data to model weights.

Developed a high-throughput REST service packaged in a public Docker container with an interactive Streamlit UI.

GitHub Repository Live Streamlit UI Docker Hub Image FastAPI Docs

Deployment

Docker Hub Public Image

Tracking

Experiment Lineage

API Serving

Low-Latency REST

Reproducibility

Deterministic Hashes

TECH:PythonScikit-LearnXGBoostDVCMLflowFastAPIDockerStreamlit

GenAI & LLMOps•Jun 2026 – Aug 2026

LIVE APPLICATION

Aurelius: Multi-Agent Research & LLMOps System

5-persona stateful LangGraph workflow with LangSmith tracing and citation verification

Autonomous multi-agent research system that coordinates 5 specialized personas to decompose complex queries, gather cross-source evidence, and generate citation-backed technical reports.

Coordinated a 5-persona stateful DAG with self-correcting review loops and automated scorecards for citation verification.

Tracked token economics and latency using LangSmith distributed tracing and Groq LPU inference acceleration.

GitHub Repository Live Streamlit UI

Multi-Agent Graph

5 Personas

Observability

LangSmith Tracing

Inference

Groq LPU

Test Suite

28 Automated Tests

TECH:LangGraphLangChainLangSmithGroq LPUPython 3.12Pydantic v2StreamlitDocker

Computer Vision•Jul 2025 – Aug 2025

EY INTERNAL PACKAGE

Inspector: Biometric Verification & Similarity Search

Fast similarity retrieval and impersonation detection at EY

In-house high-throughput biometric verification library and similarity search engine engineered at Ernst & Young (EY) to audit large-scale government examination datasets (SSC, HSSC, NHA).

Built fast similarity search across 1M+ candidate records to detect duplicate identities across nationwide exam candidates.

Implemented automated image quality filters and text matching algorithms, reducing false positive detections by ~11%.

View EY Experience

Database Scale

1M+ Candidate Records

Search Speed

Sub-Second Retrieval

Accuracy Gain

+25% Identity Detection

Distribution

Reusable Python Package

TECH:Computer VisionDeep LearningVector DatabasesSimilarity SearchPython PackagingMultiprocessing

Enterprise AI•Jan 2026 – Mar 2026

EY PRODUCTION

Candidate Intelligence Platform (CIP)

Forensic risk scoring model and LangGraph RAG investigation copilot at EY GPS Assurance

Enterprise-scale risk intelligence platform combining an ensemble candidate malpractice risk scoring engine with a LangGraph RAG copilot for automated SOP case retrieval.

Engineered ensemble risk models on multi-modal demographic and examination records for public sector audit teams.

Constructed a LangGraph RAG copilot retrieving official SOPs and historical cases for explainable, evidence-backed decision support.

View EY Experience

Orchestration

Apache Airflow DAGs

Explainability

LangGraph RAG Copilot

Lineage

Version Controlled

Backend

FastAPI Microservice

TECH:LangGraphRAGApache AirflowDVCMLflowFastAPIDockerStreamlitPython

Open Source•May 2025 – Jun 2025

LIVE ON PYPI

Logpunch: Structured Logging & Diagnostics Library

Published Python package on PyPI for production machine learning workflows

Lightweight, high-performance logging package published to PyPI to eliminate logging boilerplate, enforce Pydantic configuration schemas, and deliver module-aware stack trace diagnostics.

Published to PyPI with seamless `pip install logpunch` installation for production ML pipelines.

Enforces strict Pydantic v2 configuration validation to prevent silent pipeline logging failures.

PyPI Package

Distribution

Live PyPI Package

Command

pip install logpunch

Validation

Pydantic v2 Schemas

Telemetry

Structured JSON & ANSI

TECH:PythonPyPI PackagingPydantic v2Exception HandlingStructured Logging

RESEARCH

Research Publication

Peer-reviewed machine learning paper published in IEEE Xplore.

IEEE PEER-REVIEWED•IEEE Peer-Reviewed International Conference Publication

INDEXED IN IEEE XPLORE

Liver Disease Prediction Using Ensemble Learning

Gourav Gulia et al.•DOI: 10.1109/IC-EETA66496.2025.11548371

A clinical diagnostic predictive framework engineered for early-stage hepatic pathology detection. Evaluates comparative classification algorithms against clinical biochemical markers, utilizing statistical feature selection, cross-validation, and weighted decision voting to achieve up to 95% diagnostic accuracy.

Achieved up to 95% diagnostic accuracy, demonstrating statistically significant improvement over single-model baselines.

Calibrated high-sensitivity decision boundaries to minimize false negatives in clinical triage workflows.

Peer-reviewed and published in the IEEE Xplore digital library.

Read Paper on IEEE Xplore

Venue

IEEE Xplore

Accuracy

Up to 95%

Methodology

Weighted Ensembles

Status

Peer-Reviewed & Indexed

METHODOLOGY & STACK:PythonEnsemble LearningScikit-LearnStatistical ModelingROC-AUC CalibrationClinical Analytics

SKILLS

Technical Skills

Core technologies and frameworks used across production environments.

Machine Learning

Algorithmic modeling, feature engineering, and statistical evaluation.

VERIFIED

PythonScikit-LearnXGBoostTensorFlowModel EvaluationFeature EngineeringEnsemble Methods

Algorithmic ModelingBurnoutAI & IEEE Paper

Feature EngineeringEY Government Audits

Threshold CalibrationEY Forensic Models

Generative AI & NLP

Stateful multi-agent workflows, RAG architectures, and evaluation.

VERIFIED

LangGraphLangChainLangSmithLLMsRAGGroq LPUPydantic v2Hugging Face

Multi-Agent GraphsAnalytica & Aurelius

LLMOps TracingLangSmith Integration

Investigation CopilotsCIP Platform at EY

MLOps & Engineering

Workflow orchestration, versioned artifacts, and containerized serving.

VERIFIED

DockerMLflowDVCApache AirflowFastAPIGitHub ActionsMultiprocessing

Workflow OrchestrationEY Production Airflow

Deterministic LineageDVC Pipeline in BurnoutAI

Low-Latency APIsPublic Docker Deployments

Data & Infrastructure

High-scale vector search, relational databases, and data processing.

VERIFIED

Vector Databases (Milvus / FAISS)Similarity SearchSQLMongoDBAWSPandasNumPy

Similarity IndexingInspector Library at EY

Identity MatchingDuplicate Detection

Data ProcessingLarge-Scale Exam Audits

APPROACH

Engineering Principles

Practical engineering standards applied to production machine learning systems.

PRINCIPLE 01APPLIED IN PRODUCTION

First-Principles Problem Solving

Machine learning algorithms are grounded in statistics, linear algebra, and mathematical optimization. Understanding theoretical foundations is the only way to diagnose edge cases and build reliable production models.

Evidence:M.Sc. Data Science & IEEE Peer-Reviewed Publication

PRINCIPLE 02APPLIED IN PRODUCTION

Reproducible MLOps Pipelines

Code, data splits, and model weights must be version-controlled and reproducible. Using automated data hashes and experiment tracking guarantees full lineage from raw data to deployed artifacts.

Evidence:Automated Lineage in BurnoutAI & EY Systems

PRINCIPLE 03APPLIED IN PRODUCTION

Production-Ready Engineering

An ML model is only as valuable as its serving reliability. Building modular microservices, containerized deployments, and robust APIs ensures smooth real-world operation.

Evidence:Low-Latency APIs & Production Containers

PRINCIPLE 04APPLIED IN PRODUCTION

Evidence-Based Verification

In both biometric forensics and multi-agent GenAI, ungrounded outputs carry high risk. Automated citation validation, multi-source cross-checking, and calibrated decision boundaries ensure dependable system outputs.

Evidence:Aurelius Citation Scorecards & EY Biometric Gates

Available for ML / AI Roles

Haryana / Delhi-NCR, India

CONTACT

Get in Touch

Open to specialized Machine Learning Engineer, MLOps, and Generative AI opportunities. Available for discussions.

Download Resume (PDF)

gaurxv.gulia@gmail.com

SEND EMAIL

GITHUB

gauravgulia26

VIEW PROFILE

gauravgulia1205

CONNECT