Ishita is currently unavailable
Ishita Mukherjee
Verified Expert in Engineering
Data Engineer and Developer
Toronto, Canada
Toptal member since August 19, 2026
Bio
At Royal Bank of Canada, Ishita led the architecture and delivery of an enterprise agentic AI platform supporting 100+ business users. She is a principal engineer and architect with over 15 years of experience in data and AI/ML for financial services and healthcare. Ishita's expertise spans AWS, Databricks, and Python; she reduced infrastructure costs by 60% while at RBC.
Portfolio
RBC
PySpark, Python, Databricks, Terraform, Pandas, Spark, Scala, Kafka...
Canadian Institute for Health Information
Java, Python, PySpark, Pytest, AWS Glue, AWS CloudFormation, AWS Step Functions...
RBC Capital Markets
Scala, Spark, Spark Streaming, Apache Kafka, Python, Apache Airflow, SQL, OLAP...
Experience
- PySpark - 10 years
- Amazon Web Services (AWS) - 6 years
- Amazon SageMaker - 5 years
- Snowflake - 3 years
- Large Language Models (LLMs) - 2 years
- Databricks - 2 years
- LangGraph - 2 years
- Agentic AI Systems - 2 years
Work Experience
Principal Engineer – AI/ML, Market Risk Technologies
2022 - 2026
RBC
- Developed RAG applications utilizing Amazon Bedrock and deployed agents and knowledge bases with OpenSearch Serverless vector stores, Claude Sonnet-4 using Cross Region Inference Profiles (CRIP), and Amazon Titan Embed.
- Automated multitenant chatbot deployments across AWS consumer accounts using CDK, Step Functions, and CloudFormation, while implementing scalable ingestion pipelines through Lambda and Glue, and Python.
- Implemented memory-optimized conversational AI solutions using LangChain, custom memory buffers, and prompt chaining workflows orchestrated through Python, AWS Lambda, Step Functions, and API Gateway.
- Fine-tuned response quality through prompt engineering, temperature controls, and conversational memory management to improve user experience and response accuracy.
- Spearheaded development of a multi-agent marketing platform on AWS, utilizing Amazon Bedrock, SageMaker, Lambda, DynamoDB, and S3.
- Designed specialized copy generation, translation, and image generation agents exposed through MCP servers and orchestrated through Apache Airflow (MWAA), enabling scalable AI-driven content generation and accelerating campaign delivery timelines.
- Led architecture and development of an enterprise Agentic AI platform utilizing LangGraph, LangChain, MCP, Snowflake Cortex Analyst, and Claude Sonnet.
- Designed hierarchical multi-agent workflows incorporating planner, supervisor, and specialist agents, tool calling, relevance grading, confidence scoring, reranking, and GraphState-based orchestration.
- Implemented production observability through LangFuse.
- Integrated Snowflake Cortex Analyst, Cortex Search, and Cortex Agent for hybrid RAG implementation.
Technologies: PySpark, Python, Databricks, Terraform, Pandas, Spark, Scala, Kafka, Apache Airflow, Cloudera, Amazon S3 (AWS S3), AWS Glue, Amazon Athena, Apache Ignite, Apache Iceberg, Amazon EKS, SQL Server, SQL, Claude, Large Language Models (LLMs), LangChain, Model Context Protocol (MCP), Amazon DynamoDB, AWS Cloud Development Kit (CDK), AWS CloudFormation, LangGraph, Langfuse, Snowflake
Technical Architect – Cloud
2021 - 2022
Canadian Institute for Health Information
- Architected and led the development of HL7 FHIR-compliant pan-Canadian Organ Donation and Transplantation pipeline and data lake natively on AWS.
- Developed and deployed all aspects of inbound and outbound data pipeline processes of the new NPDUIS data platform, focusing on PySpark EMR Serverless jobs. Deployed through GitLab. Monitored through Cloudwatch logs and Cloudtrail events.
- Worked with healthcare data interoperability standards, including HL7, FHIR, and C-CDA/CCDA, to support ingestion, normalization, transformation, and downstream consumption of clinical and healthcare data.
- Used HAPI FHIR REST API deployed on AWS Fargate using Postgres DB as back end, to receive disparate health information from client systems, transform, and store in S3.
Technologies: Java, Python, PySpark, Pytest, AWS Glue, AWS CloudFormation, AWS Step Functions, AWS Lambda, Amazon S3 (AWS S3), Amazon Aurora, Amazon DynamoDB, Neo4j
Lead Data Engineer, Quant Services
2018 - 2021
RBC Capital Markets
- Architected and developed the first iteration of CM core Data Lake on-premises using Apache Spark, Scala, Hadoop, Hive, and HDFS on Cloudera Data Platform (CDP).
- Migrated on-premises data lake and Spark jobs to Azure Databricks through Azure Data Factory (ADF), leveraging Databricks SQL Warehouse for client reporting.
- Engineered a split Spark batch and Spark Streaming application in Scala, consuming from Kafka, which consumes trade data intra-day and end of day from Front Office.
- Migrated apps hosted on Docker Swarm to Azure Kubernetes Service.
- Architected and helped develop a robust back-end service in Java 11/Spring Boot with a REST layer to allow uploading large data files from front-office clients into HDFS, complete with versions, reruns, and partial reruns.
Technologies: Scala, Spark, Spark Streaming, Apache Kafka, Python, Apache Airflow, SQL, OLAP, Docker, Kubernetes, YAML, Spring, Java, Apache Druid, Azure Databricks
Senior Data Engineer, Wealth Data Systems
2016 - 2018
TD Bank Group
- Led a team of 8 developers to build 3 applications for various lines of business across the client reporting team; each application processed 3-4 gigabytes of raw data, generating batch client statements in XML format using Apache Spark on Hadoop.
- Wrote unit tests in Flatspec with ScalaTest, Scala Mock, and JUnit.
- Migrated high-performance big data projects from Cloudera to Azure using Azure Data Factory, Azure Data Lake Storage (ADLS), and Azure Databricks Lakehouse built on Medallion architecture.
- Ingested streaming trading data using Spark streaming and Apache Kafka. Configured Spark and YARN configurations for optimum performance.
- Wrote a derivative recommendation engine for investment banking clients using Spark, Scala, and supervised machine learning algorithms of MLLib.
Technologies: Databricks, Blob Storage, Scala, Hadoop, Apache Hive, HDFS, Yarn, Cassandra, Apache Kafka, Spark Streaming, SQL Server, Elasticsearch
Senior Software Engineer
2015 - 2016
Autotrader
- Developed multithreaded RESTful service APIs with message queuing (Solace JMS) using core Java and Jax-RS, deployed on Azure VMs, hit with 4 million calls per day.
- Wrote custom authentication middleware that implements OAuth standards. Created self-deleting refresh tokens stored in Azure Redis distributed cache.
- Wrote high-performance Hibernate Query Language queries to a DAO layer optimized for ultra-fast response times.
- Wrote JMS-based user activity message queuing integrated with the cloud database.
Technologies: Java, Spring MVC, JavaScript, Node.js, Redis, SQL Server
Senior Software Engineer
2014 - 2015
FCT
- Handled developing and maintaining an upgraded and robust, SOAP-based, back-end WCF service for all title-related operations using .NET and C#. Service was backed by Entity Framework as the ORM.
- Implemented Lambda expressions and lazy loading to optimize ORM query performance.
- Built a service layer using dependency injection and Inversion of Control (IoC), Log4net for logging, Nunit for unit tests, and atomic transactions to ensure no stale read/write.
- Wrote stored procedures to query and view a large quantity of data from the Oracle database. Designed and optimized WPF controls to work with the data.
Technologies: .NET, C#, WCF Web Services, Entity Framework
Senior Full-stack Engineer
2014 - 2014
Finastra
- Developed reusable, lightweight, proprietary HTML5 web control library with fluent interface method chaining. The controls ran on JavaScript frameworks like AngularJS, Knockout.js, JQWidgets, and jQuery for use in Jakarta EE and .NET applications.
- Built a single-page web application (SPA) with Node.js, Express.js, and Redis key-value store, offered as SaaS to financial clients.
- Developed back-end services on .NET, C#, and REST API.
Technologies: Angular
Full-stack Engineer
2013 - 2014
Citi
- Developed the algorithm and HTML5-based UI for a dynamic data display system for Citi’s commercial mortgage financing platform in Spring framework using Java.
- Implemented Aspect-Oriented Programming using Spring AOP for logging (Log4j) and application monitoring (JMX). Built a service layer using Spring Singleton, used Spring Annotation and Spring Transaction modules to manage distributed transactions.
- Wrote highly performant queries in Hibernate Query Language to fetch data from the Oracle database. Consumed messages from legacy systems using JMS (ActiveMQ).
- Designed and developed a high-throughput Java RESTful web service using JAX-RS, which was consumed by multiple real-time web and desktop applications throughout Citi’s capital market platforms.
Technologies: Spring, Java
Software Developer – Co-op
2012 - 2013
Schneider Electric
- Developed a bulk data import/export system for enterprise software using Java Spring MVC, Silverlight, and .NET WCF host tied to a Jakarta EE back end, which accessed the database through JDBC (implementing ODBC) API.
- Built a C# Windows Service that processes on demand 2-way communications between a file management API and a 3rd-party API.
- Implemented Tree traversal algorithms to find and display forward-only hierarchies for internal tools.
- Designed a relational database management system to find and store decompressed XML data in an Oracle database.
Technologies: .NET, Java, C#
Project History
Agentic Chatbot
Led the architecture and development of an enterprise Agentic AI platform utilizing LangGraph, LangChain, MCP, Snowflake Cortex Analyst, and Claude Sonnet.
I designed hierarchical multi-agent workflows incorporating planner, supervisor, and specialist agents, tool calling, relevance grading, confidence scoring, reranking, and GraphState-based orchestration. I also implemented production observability through LangFuse and integrated Snowflake Cortex Analyst, Cortex Search, and Cortex Agent for hybrid RAG implementation.
I designed hierarchical multi-agent workflows incorporating planner, supervisor, and specialist agents, tool calling, relevance grading, confidence scoring, reranking, and GraphState-based orchestration. I also implemented production observability through LangFuse and integrated Snowflake Cortex Analyst, Cortex Search, and Cortex Agent for hybrid RAG implementation.
Education
2010 - 2012
Master's Degree in Electrical Engineering
University of Victoria - Victoria, BC, Canada
Certifications
MAY 2024 - PRESENT
Professional Certificate in Artificial Intelligence
Stanford School of Engineering
DECEMBER 2020 - DECEMBER 2023
AWS Certified Data Analytics – Specialty
Amazon Web Services (AWS)
JUNE 2012 - PRESENT
Microsoft Certified Technology Specialist (MCTS): .NET 4, Web Applications
Microsoft
Skills
Libraries/APIs
PySpark, Spark Streaming, REST APIs, Pandas, Node.js, Apache Lucene, Entity Framework
Tools
Cloudera, AWS Glue, Amazon Athena, Apache Ignite, Apache Iceberg, Amazon EKS, Claude, AWS Cloud Development Kit (CDK), AWS CloudFormation, Pytest, AWS Step Functions, Apache Airflow, Amazon SageMaker, AWS IAM, Terraform, Apache Druid, Oozie
Languages
Python, Scala, SQL, Snowflake, Java, YAML, C#, JavaScript
Frameworks
Spark, LangGraph, Spring, Hadoop, Yarn, Apache Spark, Spark Structured Streaming, Spring Boot, Spring MVC, .NET, Angular
Paradigms
Model Context Protocol (MCP), OLAP
Platforms
Langfuse, AWS Lambda, Apache Kafka, Docker, Kubernetes, LangSmith, Amazon Web Services (AWS), Databricks, Azure
Storage
Amazon S3 (AWS S3), Amazon DynamoDB, Apache Hive, HDFS, Elasticsearch, Redis, NoSQL, Redshift, PostgreSQL, Amazon Aurora, Neo4j, Cassandra, MongoDB
Other
SQL Server, Large Language Models (LLMs), LangChain, Blob Storage, Agentic AI Systems, Agentic AI, Kafka, Azure Databricks, Linear Regression, Logistic Regression, Neural Networks, Web Applications, Quantum Computing, Amazon Kinesis, Hbase, MVC Frameworks, WCF Web Services
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring