Saswata Roy

About Saswata Roy

I’m a Technical Lead and Machine Learning Engineer with 10+ years of experience designing intelligent, scalable systems across cloud, data engineering, and applied GenAI. My work spans consulting, smart city analytics, and enterprise product innovation—where I bridge ML engineering, LLM systems design, and pragmatic product thinking.I specialize in architecting LLM-powered automation, data platforms, and cloud-native pipelines using Azure (ADF, Databricks, ADLS), Airflow, Kubernetes, and FastAPI—serving 20+ global clients across industries.Recently, my focus has been on leading end-to-end GenAI platform engineering, including: • AI Rule Translator → Converts natural-language business rules into validated Spark SQL/UDF expressions using LangGraph + Azure OpenAI, with strict guardrails and human-in-the-loop review. • DPD (Data Process Document) Generation Agent → Extracts structured metadata from unstructured client PDFs/Excels, summarizes, validates, and generates standardized onboarding documents for rapid client enablement. • Config Generator Pipeline → Transforms DPD outputs into client-specific ingestion, metadata, DQ, transformation, and scheduling configs, automatically stored in ADLS for “plug-and-play” activation on the ADA platform. • NLP KPI Agent → Enables natural-language querying of PostgreSQL and automatic chart/table generation using text-to-SQL with LLM scoring and fallback strategies. • Expertise in async architectures, streaming evaluation, workflow orchestration, stateful agent design, and secure cloud integration.I’m passionate about building AI-first, modular, auditable systems that reduce operational effort, accelerate onboarding, and make complex data processes accessible to non-technical users. I thrive at the intersection of ML engineering, software design, and product strategy, driving ideas from prototype to production at enterprise scale.

Experience

1. Senior Technical Lead

   EY

   May 2021 — Present · Bengaluru, IN

   As a Lead Engineer for EY’s ADA data automation platform, I build production-grade GenAI systems, rule engines, and cloud-native ETL pipelines used to onboard and operate data flows for 10+ global clients. My work spans GenAI agents, onboarding automation, and scalable Databricks/ADF architectures that standardize and accelerate client delivery.Key ContributionsGenAI & Automation • Built the AI Rule Translator (LangGraph + Azure OpenAI) converting natural-language rules into validated Spark SQL/UDF expressions. • Developed the DPD Generation Agent to extract and summarize client metadata from PDFs/Excels and auto-generate onboarding documents. • Added a Config Generator that turns DPD outputs into client-specific ADA configs (ingestion, metadata, DQ, transformations, schema mapping, scheduling) stored in ADLS for plug-and-play activation. • Implemented interrupt/resume flows, async execution, checkpointing, and guardrails for multi-step LLM pipelines.ETL & Platform Engineering • Architected modular ingestion → validation → transformation pipelines using Databricks, ADF, ADLS, supporting CSV, Excel, JSON, ZIP, SAP CDS, and API sources. • Designed the Business Rule Engine for schema mapping and metadata validation across clients. • Optimized ADF–ADB communication, integrity checks, dedupe logic, and Delta Lake storage patterns to reduce runtime and cost. • Built reusable notebooks and parameterized workflows enabling rapid, low-effort client onboarding.Impact • Reduced onboarding effort from weeks to hours through automated DPD + config generation. • Standardized ingestion and rule processing across 10+ enterprise clients. • Cut manual rule-writing by 80%+ via AI-assisted rule translation. • Improved pipeline performance and cost efficiency through architectural redesign.

Education

• West Bengal University of Technology, Kolkata

  Bachelor of Technology (B.Tech.), electrical engineering

  2011 — 2015

• B.P. Poddar Institute Of Management and Technology

  B.tech, Electrical engineering

  2011 — 2015

• Kanailal Vidyamandir

  Science

  1999 — 2011

• Udacity

  NanoDegree, Data Engineer

Skills

• Sql
• Linux
• Jpa
• Mysql
• Customer Service
• Persistence
• Sqlite
• Javascript
• Neo4j
• Mobile Devices
• Core Java
• Electrical Engineering
• Hana
• J2ee Web Services
• Php
• Image Manipulation
• English
• Python
• Javaserver Pages (Jsp)
• Xml
• Post Processing
• Sap Ui5
• Photography
• Dbms
• Android
• C
• Java
• Android Development
• Sap Hana
• Android Sdk
• Servlets
• Windows
• Eclipse