PKGPrasanta Kumar Gorai
Let's talk ↗ LinkedIn ↗ GitHub ↗ Download résumé ↓

Available for ambitious systems

Backend & Cloud Engineer building scalable enterprise systems.

I’m a Backend Team Lead and AWS Certified Solutions Architect with 10+ years of experience building Python services, cloud infrastructure, enterprise integrations, and AI-powered systems.

I specialize in turning complex business workflows into reliable, scalable software.

Explore selected work ↓
10+ years buildingAWS certified architect∞ curiosity for better systems
LIVE SYSTEM MAP01 / 04
AI Agent
FastAPI
AWS Cloud
Data / APIs
Building resilient digital infrastructure
10+years of
engineering
PythonCloud ArchitectureAPI IntegrationAgentic AISystem DesignPythonCloud ArchitectureAPI IntegrationAgentic AISystem Design
01

Selected work

Built for the hard parts.

Enterprise-grade applications where integrations, reliability, and speed all matter.

01
STRATAVAR / CISCO 360

Lead intelligence,
from signal to action.

React + ViteFastAPIAWS
Salesforce
Connected
BUILD / 2026

01 / Enterprise platform

Cisco 360 LeadGen

A full-stack estimate and lead system, secured with passwordless access and connected to Salesforce.

Architecture — request path

Architecture — daily GDPR compliance path

Problem

Sales teams needed a fast way to generate Cisco estimates and leads without manual re-entry into Salesforce or shared, insecure logins.

My role

Designed and built the serverless AWS stack end-to-end — the Terraform infrastructure, the FastAPI/Lambda backend, the Cognito auth integration, and the CloudFront/API Gateway routing.

Technology

React, Vite, FastAPI on AWS Lambda (Mangum), API Gateway, CloudFront, S3, Amazon Cognito, RDS MySQL (IAM auth), EventBridge, Terraform, Salesforce APIs.

Key challenges

Serving a SPA and a serverless API from one CloudFront domain without exposing the raw API Gateway URL, and connecting Lambda to RDS without ever storing a static database password.

Solution

Fronted both the S3-hosted SPA and API Gateway with a single CloudFront distribution, protected by a shared-secret header the Lambda validates; used IAM database authentication for RDS instead of a static password; and isolated a separate EventBridge-scheduled Lambda to handle GDPR purges outside the VPC.

Impact

One public hostname serves the whole product, the API is unreachable except through CloudFront, there's no long-lived database credential anywhere, and GDPR retention is enforced automatically every day.

02
MCP
Claude
Quoting
Rebates

02 / Agent infrastructure

AI-ready internal tools

An MCP server exposing secure quoting and rebate-calculation capabilities to AI agents.

Architecture

Problem

Internal quoting and rebate calculations lived behind manual workflows and internal-only APIs that an AI agent couldn't safely call.

My role

Designed and built the MCP server, defined the tool schemas and permissions, and wired Bedrock-based routing between agent intent and the right tool.

Technology

MCP (Model Context Protocol), AWS Bedrock, Claude tool-use, Python/FastAPI, internal pricing APIs.

Key challenges

Exposing internal pricing and rebate logic to an agent safely — scoping exactly what each tool call can see and do — without opening the underlying systems directly.

Solution

Wrapped quoting and rebate calculations behind narrowly-scoped MCP tools, so an agent gets exactly the capability it needs and nothing more.

Impact

Internal pricing and rebate logic became something an AI agent can use directly and safely, instead of staying locked behind manual, human-only workflows.

03
STRUCTURED
EXTRACTION SMART

03 / Intelligent automation

Document intelligence pipeline

Schema-driven PDF quote parsing with model escalation and self-correction built in.

Architecture

Problem

Vendor quotes arrive as unstructured PDFs in inconsistent formats, making reliable structured-data extraction hard to trust.

My role

Designed the extraction pipeline end-to-end, including the schema constraints, the self-correction loop, and the escalation logic.

Technology

Python, LLM-based extraction, schema validation, model escalation logic.

Key challenges

Getting consistent, trustworthy structured output from documents that vary widely in layout and quality, without a human reviewing every one.

Solution

Constrained extraction to a strict schema, added a self-correction pass that catches its own low-confidence output, and escalated only the genuinely hard cases to a stronger model.

Impact

Turned inconsistent vendor PDFs into structured, trustworthy quote data without a human checking every extraction.

Also shipped

Architecture

Problem

Long-running jobs triggered from Salesforce needed to run without blocking Salesforce or getting lost if something failed.

My role

Built the FastAPI service and the SQS-based queuing so Salesforce-triggered work runs reliably in the background.

Technology

FastAPI, Amazon SQS, Coralogix.

Key challenges

Making sure jobs triggered from Salesforce don't get dropped or double-processed, with enough visibility to know when something actually failed.

Solution

Queued every job through SQS instead of processing inline, and instrumented the path with Coralogix so failures are visible instead of silent.

Impact

Salesforce-triggered work runs reliably in the background with clear visibility when something goes wrong.

Architecture

Problem

Cisco, TD SYNNEX, and Ingram Micro each ship quotes in their own Excel/PDF format, making normalization a constant maintenance burden.

My role

Designed the configuration-first parsing approach and built the REST integrations into each distributor and Salesforce.

Technology

Python, Excel/PDF parsing, REST APIs, Salesforce REST, Bulk, and Metadata APIs.

Key challenges

A new or changed vendor format used to mean new code — the goal was to make that a configuration change instead.

Solution

Built a configuration-first parser where each vendor's format is described in config, not hard-coded, so format changes don't require a redeploy.

Impact

Distributor format changes became a config update instead of an engineering task, and the same pipeline scaled across three major distributors.

Architecture

Problem

AWS spend and infrastructure had grown organically without a deliberate cost or performance strategy.

My role

Led the audit and modernization work and codified the resulting infrastructure as CloudFormation templates.

Technology

AWS EC2, RDS, Graviton, Reserved Instances, Savings Plans, CloudFormation.

Key challenges

Cutting cost without hurting performance or reliability meant understanding real usage patterns before committing to RI or Savings Plan terms.

Solution

Matched workloads to Graviton where compatible, committed to RIs/Savings Plans based on actual usage, and moved the result into CloudFormation.

Impact

Lowered AWS infrastructure cost while keeping the infrastructure reproducible and version-controlled.

Architecture

Problem

Every Salesforce quote format needed its own export logic, coupling document layout to application code and making a new or modified format a development task instead of a configuration change.

My role

Designed and built the template-driven export engine — the JSON data template schema, the Excel layout template renderer, and the pipeline connecting Salesforce data to dynamic document generation.

Technology

Python, Salesforce SOQL, JSON data templates, Excel-based layout templates, HTML-formatted content rendering.

Key challenges

Building one rendering engine generic enough to handle arbitrarily different quote structures and layouts, while dynamically expanding rows and columns without breaking the template's formatting.

Solution

Split the engine into two independent templates — a JSON data template describing the required Salesforce queries and data shape, and an Excel layout template describing presentation — so the core engine only interprets templates and never contains quote-specific logic. Rendering dynamically expands rows and columns from the data while preserving the template's formatting, including embedded HTML-formatted content.

Impact

New or modified quote formats are introduced by creating or editing templates rather than writing new export code, keeping the core engine unchanged as formats evolve.

02

What I bring

Depth where it
counts. Range
where it matters.

01

Backend systems

Python, FastAPI, Django, asynchronous jobs, REST APIs, and architecture designed to scale.

↗
02

Cloud engineering

AWS infrastructure across EC2, RDS, Lambda, SQS, S3, Cognito, CloudFormation, and more.

↗
03

Systems integration

Reliable connections across vendor platforms, distributors, Salesforce, document formats, and internal tools.

↗
04

AI-native workflows

Bedrock, Claude tool-use, MCP, and practical agentic pipelines that make work move faster.

↗

Tech stack

Backend

PythonFastAPIREST APIsAsyncIO

Cloud

AWSEC2LambdaAPI GatewayS3SQSRDSCognitoCloudFormation

Data

MySQL

Integration

SalesforceREST APIsGraphQLOAuthWebhooks

AI

AWS BedrockClaudeMCPLLM Tool CallingDocument Intelligence

DevOps

DockerJenkinsTerraformInfrastructure as CodeObservability
03

How I build

Principles I actually build by.

Not values-page fluff — the defaults I reach for on every system I touch.

01

Reliability first

Systems should degrade gracefully and recover automatically.

02

Simple architecture

Avoid unnecessary complexity until scale actually requires it.

03

Observable by default

Logs, metrics, and traces are part of the system — not an afterthought.

04

Automation over repetition

If a workflow happens repeatedly, it should probably be automated.

05

AI where it creates leverage

Use AI to improve workflows, not simply because AI is fashionable.

04

About Prasanta

I like the work that sits between an ambitious idea and a system that can actually carry it.

From leading backend teams to designing cloud infrastructure and untangling high-stakes integrations, I bring calm, clear thinking to complex technical challenges. I’m open to backend and cloud leadership roles, agentic AI collaborations, and difficult integration problems.

Start a conversation ↗
AWSCertified Solutions Architect
Associate

A decade in motion

2026 — now

Digetful Innovations

Backend Team LeadLeading enterprise quoting, vendor-integration, and agentic AI work for StrataVAR.

2020 — 2026

221B Baker Street

Lead Software EngineerBuilt AWS and Python platforms for enterprise quoting, Salesforce, and distributor APIs.

2018 — 2019

HCL Technologies

Lead EngineerDelivered a Dockerized real-time image analytics API with WebSocket notifications.

2015 — 2018

IDEMIA

Software EngineerStrengthened Python build systems, automation, and developer tooling.
02/2026 — PRESENT

Leading backend systems that connect people, platforms, and AI.

  • Built Cisco 360 LeadGen with React, FastAPI, Cognito passwordless authentication, RDS MySQL, Salesforce, and an AWS delivery/security layer.
  • Designed MCP tools and Bedrock-based agentic routing across vendor systems and internal pricing engines.
  • Built reliable document intelligence with schema-constrained extraction, self-correction, and model escalation.
01/2020 — 01/2026

Made complex quoting and distributor workflows dependable.

  • Integrated Cisco, TD SYNNEX, and Ingram Micro securely with REST APIs and Salesforce REST, Bulk, and Metadata APIs.
  • Created configurable Excel/PDF parsing and quote-generation workflows for diverse distributor formats.
  • Designed cloud infrastructure across EC2, RDS, Lambda, API Gateway, SQS, S3, and CloudFormation.
08/2018 — 12/2019

Shipped real-time analytics services.

  • Designed a REST API for image analytics with live WebSocket notifications.
  • Used multithreading to process frames from multiple RTSP cameras and connected analytics to the API layer.
  • Containerized the solution with Docker and built C/Java coding-guideline validation tooling.
05/2015 — 08/2018

Built the engineering foundations.

  • Enhanced an internal Python build system, including its DSL, size calculations, and dependency graphs.
  • Automated workflows with Python, Perforce, and Git; developed with Django, Tkinter, and wxPython.
  • Adapted xlsxwriter, openpyxl, Sphinx, pandas, and NumPy to solve project-specific needs.

Have a complex problem?

Let's make it
work beautifully.

prasanta4nit@gmail.com ↗