Available for opportunities

Pooja
Tumma

specializing inMulti-Agent Systems

Building production-grade AI systems, agentic workflows, and scalable backend infrastructure. From multi-agent orchestration to distributed RAG pipelines — I ship intelligent systems that work reliably at scale.

2+
Years shipping AI
6
Enterprise deployments
58%
Latency reduction
SF Certifications
Mumbai, India
terminal — pooja@ai-systems
LangGraphRAGPythonAWS LambdaRedisSupabaseAgentforce
scroll
Systems built

Featured Projects

Production-grade AI systems, not hobby projects. Each one solves a real problem with measurable outcomes.

StratumAI

FlagshipLive

Multi-Agent Intelligence Platform

A parallel multi-agent orchestration system that runs specialized AI agents concurrently to deliver comprehensive intelligence reports in real time. Built with LangGraph StateGraph, distributed Redis caching, and Supabase-backed persistence — engineered for production reliability.

58%
Latency reduction
vs sequential baseline
3x
Agents in parallel
concurrent execution
6h
Cache TTL
Redis / Upstash
100%
Persistence
restart survivable
LangGraphLangChainPythonMistral AITavilyUpstash RedisSupabaseStreamlitThreadPoolExecutor

Generative AI Chatbot

RAGComplete

RAG Pipeline with Semantic Retrieval

A Retrieval-Augmented Generation pipeline that dramatically reduces hallucinations through semantic chunk retrieval, reranking, and context-aware prompt engineering — with an interactive Gradio interface.

38%
Hallucination reduction
manual eval, 100Q set
256
Chunk size
tokens (tuned from 512)
Semantic
Retrieval
with reranking
Gradio
Interface
real-time Q&A
PythonLangChainOpenAIChromaDBFAISSGradioVector Embeddings

NEWSIN

ResearchComplete

Real-Time News Intelligence

An NLP pipeline that processes 100+ news articles per minute using TextRank extractive summarization and VADER sentiment analysis. Published research, recognized in Top 1% at institute showcase.

100+
Articles/minute
real-time pipeline
Top 1%
Recognition
institute showcase
IJRASET
Published
Vol 10, 2022
2
NLP methods
TextRank + VADER
PythonNLPTextRankVADER SentimentRSSBeautifulSoup
Work history

Experience

2 years designing, shipping, and owning production systems — from AI agent workflows to distributed enterprise integrations.

E

AI & Software Engineer

Exceller Tech·Salesforce Consulting + Product Development
June 2024 – Present
Mumbai, India
Agentforce AI Workflows
AgentforceAI AutomationSalesforce

Architected and deployed production Salesforce Agentforce AI agent workflows for enterprise CRM clients — translating ambiguous requirements into shipped AI features reducing manual process overhead by 40%.

Distributed Microservices
AWS LambdaNode.jsMicroservices

Sole technical owner of 6 enterprise integration deployments using Node.js + AWS Lambda — processing high-volume async messaging workloads with fault-tolerant, zero-downtime operations.

Event-Driven Architecture
Async FrameworksEvent-DrivenApex

Designed Apex Queueable/Batch async frameworks and event-driven pipelines eliminating 40% manual operational overhead — idempotent, resilient under partial failure.

Fault-Tolerant Integrations
REST APIsFault ToleranceEncryption

Built payment and CTI integrations using RESTful APIs with strong encryption, retry mechanisms, and circuit-breaker patterns — stable under ambiguous edge cases and network partitions.

Enterprise SaaS
SaaSDistributed SystemsMulti-tenant

Contributed to enterprise document generation SaaS — shipped backend modules for distributed data processing serving multi-tenant enterprise clients at scale.

Outstanding Performance Award, 2024
Tools & Technologies

Tech Stack

A systems-first stack built around production AI: agent orchestration, vector retrieval, async backends, and cloud infrastructure.

AI / LLM
LangGraphLangChainLlamaIndexRAG PipelinesMulti-Agent SystemsPrompt EngineeringOpenAI APIMistral AIAgentforceTavilyRAGAS
Vector & Retrieval
ChromaDBFAISSUpstash RedisVector EmbeddingsSemantic SearchReranking
Backend
PythonNode.jsFastAPIREST APIsMicroservicesEvent-DrivenAsync SystemsWebhooks
Cloud & DevOps
AWS LambdaAWS S3DockerCI/CDGitHub Actions
Databases
SupabasePostgreSQLMongoDBRedis
Languages
PythonJavaScript (ES6+)TypeScriptJavaApexSQLC++
System Design

Architecture

The systems I design go beyond LLM calls. Parallel execution, fault tolerance, distributed state — production patterns from day one.

LangGraph StateGraph with parallel ThreadPoolExecutor dispatch across 3 specialized agents

User QueryRedisCacheLangGraphOrchestratorMarketAgentTechnicalAgentSentimentAgentTavilySearchResultAggregatorSupabaseStoreIntelligenceReport
Input/Output
Core Logic
Agent
Infrastructure
External Tool
Recognition

Achievements

Credentials, recognitions, and milestones across engineering, AI research, and leadership.

Certification

Agentforce Specialist

Salesforce2024
Certification

Data Cloud Consultant

Salesforce2024
Certification

Getting Started with Deep Learning

NVIDIA2023
Certification

Introduction to Generative AI

AWS Educate2023
In Progress

AI-First Product Management

Airtribe2025
Award

Outstanding Performance Award

Exceller Tech2024
Published

Research Publication

IJRASET, Vol 102022
Education

B.E. AI/ML Honors · CGPA 9.65

APSIT Mumbai2024
Get in touch

Let's build something
intelligent together.

Open to AI Engineering, Applied AI, and GenAI opportunities.

Based in Mumbai · Open to remote