Hi, I'm
Sourav Halder
System Metrics & AI Operations
Operational telemetry derived from enterprise AWS Bedrock Agent Core deployments, GenAI automation platforms, and cloud orchestration.
-0%
L1 Cognitive Load
Omnichannel support system on AWS Bedrock Agent Core with Supervisor & sub-agent clusters, multi-source RAG, and Confident AI evaluation, cutting repetitive L1 cognitive load by 70%.
-0%
Turnaround Time Cut
Spearheaded an n8n and LLM-driven release communication platform broadcasting channel-native layouts (HTML emails, Teams Adaptive Cards) with HITL tone selection, cutting turnaround time by 85%.
-0%
MTTR Improvement
Streamlined incident management via ServiceNow API for pipeline monitoring, automatic ticket creation, and intelligent team assignment, decreasing mean time to resolution by 40%.
0.0%
Operational SLA
Event-driven pipeline orchestration (Lambda, EventBridge, S3) with programmatic retry logic and validation checks, eliminating synchronization failures by 85% with 99.5% reliability.
Expertise & Services
Bridging the gap between raw data systems, machine learning models, and secure cloud orchestration.
Agentic AI & GenAI Systems
Architecting autonomous multi-agent workflows, Supervisor Agent patterns, and Model Context Protocol (MCP) tool delegation. Specializing in LangChain, LangGraph state machines, n8n automations, Portkey AI gateway, and Confident AI evaluations.
- Supervisor Multi-Agent Patterns
- Model Context Protocol (MCP)
- AWS Bedrock Knowledge Bases
- n8n Workflow Automation
- Portkey Gateway & Evals
AI/ML & Data Science
Training predictive models, deep neural networks, and anomaly detection pipelines. Transforming telemetry into actionable machine intelligence with Optuna and MLflow.
- Predictive ML & XGBoost Ensembles
- Hyperparameter Tuning (Optuna)
- Neural Network Modeling
- Feature Engineering & MLflow
AWS Cloud Architecture
Designing enterprise-grade, cost-optimized cloud architectures on AWS and modern Lakehouses. 4x AWS Certified & 3x Databricks Certified.
- AWS Serverless & Containers
- Medallion Lakehouse (Iceberg / Delta)
- Multi-AZ Cloud Resiliency
- Managed RDS, DynamoDB & Redshift
Professional Experience
Detailed operational domains and systems implemented during my consulting career.
Senior Associate Consultant
Infosys LimitedServing as a Senior Associate Consultant at Infosys Limited, architecting Generative AI PaaS platforms, AWS Bedrock Agent Core omnichannel systems, production RAG pipelines, and high-reliability data lakehouse infrastructure.
Key Contributions & Achievements
- Engineered and maintained 50+ production ETL pipelines using Python, SQL, and AWS Glue, orchestrating data extraction from multi-source SAP ERP, SQL Server, and REST APIs into Redshift data warehouse supporting 14 data marts, ensuring timely data availability for business dashboards and analytics teams.
- Implemented robust pipeline orchestration with event-driven and schedule-based triggers (Lambda, EventBridge, S3), incorporating programmatic retry logic and data validation checks that eliminated synchronization failures by 85%, resolved data mismatches, and achieved 99.5% pipeline reliability.
- Re-architected data extraction pipeline by migrating from direct SQL database to Apache Iceberg-based live layer with incremental delta loads, achieving 93.75% reduction in data extraction and eliminating performance bottlenecks caused by high-volume ERP replication traffic.
- Streamlined incident management by integrating ServiceNow API for pipeline monitoring, enabling automatic ticket creation and intelligent team assignment for job failures; decreased mean time to resolution (MTTR) by 40% and eliminated manual triaging overhead.
- Architected a Generative AI PaaS integrating Portkey Gateway (routing, rate limiting, serving, usage tracking) and AWS Bedrock for centralized RAG capabilities; secured multi-tenant access via Active Directory (AD) and SailPoint to allow teams to securely ingest and query proprietary data.
- Automated enterprise workflows using n8n, engineering a Jira ticketing agent that queries Bedrock Knowledge Bases to auto-suggest resolutions; designed LLM-generated stakeholder reports and a sync workflow from SharePoint to Confluence/Web.
- Spearheaded an n8n and LLM-driven release communication platform, broadcasting channel-native layouts (HTML emails, Teams Adaptive Cards) via self-service forms and CI/CD webhooks; integrated HITL tone selection and audit logs, cutting turnaround time by 85%.
- Delivered an omnichannel support system on AWS Bedrock Agent Core, deploying an orchestrator across specialized Supervisor and sub-agent clusters; engineered multi-source RAG with confidence-scored fallback web search, Jira escalation, and Confident AI evaluation, cutting repetitive L1 cognitive load by 70%.
Education Records
Academic qualifications, specialized engineering domains, and formal credentials.
Technical Skills
Organized the way I think about engineering and automation — directly inside a simulated code editor.
Featured Projects
Architecture-first implementations focused on cloud automation, data pipelines, and machine learning structures.
Multi-Agent Enterprise System
Supervisor Pattern + MCP Tool Delegation + HITL
Autonomous multi-agent orchestration framework utilizing the Supervisor Agent pattern and Model Context Protocol (MCP) to dynamically delegate tasks across specialized domain agents with stateful Human-in-the-Loop validation.
Production Agentic RAG Platform
LangGraph + Stateful PostgreSQL Checkpoint + LangSmith
Enterprise agentic RAG platform engineered with LangGraph and LangChain, featuring stateful PostgreSQL checkpointers for session memory, dynamic tool execution, hybrid vector retrieval, and full-stack LangSmith observability.
SeismicSense: Multi-Source Geophysical Data Pipeline
Type 1 & Type 2 SCD + Incremental Delta Load + PostgreSQL
Production-grade ETL pipeline extracting telemetry from USGS Earthquakes, NOAA CO-OPS Tides, and NASA POWER APIs into a normalized PostgreSQL/PostGIS data warehouse with automated SCD Type 1/2 merges and run observability.
Network Security System: End-to-End MLOps Pipeline
ETL Pipeline + MLOps + XGBoost & Optuna + AWS Docker
Production-grade MLOps network threat detection system featuring automated MongoDB ETL ingestion, SMOTE class balancing, Optuna-tuned ensemble models, MLflow Champion/Challenger registry, and containerized AWS deployment with Evidently AI drift monitoring.
Kidney Disease Classification & MLOps Pipeline
EfficientNetV2-B0 + Transfer Learning + DVC + MLflow + AWS EC2
Production-grade deep learning pipeline classifying kidney tumors from CT scans using fine-tuned EfficientNetV2-B0, automated DVC reproducibility, and MLflow Model Registry deployed on AWS EC2.
Data Science Playground: Interactive ML Sandbox
Real-Time Hyperparameter Tuning & Decision Boundaries
Interactive Streamlit platform for exploring machine learning algorithms hands-on, dynamically adjusting hyperparameters to visualize 2D decision boundaries, ROC curves, and performance metrics.
AWS Three-Tier Web Architecture
Auto Scaled Load Balanced Infrastructure
Utilized a public Application Load Balancer to route client traffic to auto-scaling Nginx web servers, communicating with internal application servers and RDS databases.
Spotify Big Data Analytics Pipeline
Serverless Ingest & Aggregation ETL
Built a serverless data pipeline to extract, transform, and load Spotify top songs data into a visual analysis dashboard using AWS serverless analytics services.
DynamoDB Real-Time Audit Table
NoSQL Database Mutation Capturing
Designed an automated audit trail system for DynamoDB databases, tracking data mutations in real-time for compliance and operational analytics.
WildRydes Serverless App
Complete Serverless App Ingestion
Built an end-to-end serverless ride-sharing web application leveraging AWS serverless services to manage user authentication, ride requests, and backend processing.
Medallion Data Warehouse Schema
Layered Relational Database Design
A comprehensive data warehousing solution, cataloging and exploring relational datasets using SQL scripts to structure Bronze, Silver, and Gold layers.
Achievements & Certifications
Professional certifications and recognition earned throughout my career. Click any badge to view credentials.
