Rahul Reddy Talatala
$0M+Overcharge exposure
$0.00MCost avoidance
0%Debug time reduced

Rahul Reddy Talatala

Iautomate
the complex.
ImakeAI
work at scale.

GenAI Engineer building production-grade LLM systems, multi-agent pipelines, and AI infrastructure that ships with real, measurable impact.

Download CV
About

Building AI that ships,
not just demos.

I'm a GenAI Engineer who turns cutting-edge AI research into systems that actually ship. I've automated millions of dollars in operational waste, fine-tuned LLMs for production inference, and built agentic pipelines that work at enterprise scale, not just in notebooks. I care about real impact over cool demos.

Whether it's orchestrating multi-agent pipelines, fine-tuning LLMs for inference efficiency, or building telecom-scale RAG systems, I live at the intersection of research and production. Currently at Verizon, previously at Apple.

🧠

Agentic Systems

LangGraph, MCP, multi-agent orchestration at enterprise scale

📊

RAG & Knowledge Graphs

Hybrid Graph RAG with Neo4j, Elastic, and pgvector

⚡

MLOps & LLMOps

NVIDIA Triton, LoRA fine-tuning, LangFuse observability

☁️

Cloud Infrastructure

AWS, GCP, Azure: Kubernetes, Terraform, CI/CD at scale

Experience

Where I've shipped.

GenAI Engineer

Current
Infinite Computer Solutions·Client: Verizon
Aug 2025 – PresentTampa, FL
  • Built a five-stage document-audit pipeline with Gemini 2.5 Flash and a 70-rule engine that matches invoice line items to purchase orders via hybrid search, surfacing $114–140M in annual overcharge exposure across 500–2,000 capital projects.
  • Fine-tuned Qwen2.5-3B with LoRA to mask sensitive financial fields on-prem before any data reaches a frontier LLM, then used DSPy to optimize extraction prompts — lifting structured-field accuracy from 74% to 91%.
  • Solved the cold-start problem for a 346-table text-to-SQL agent by building an offline pipeline that builds a 1,536-dim ChromaDB vector store and a 17,772-edge NetworkX join graph, enabling accurate retrieval from day one across 11.8M+ rows.
  • Designed a multi-agent FiOS planning system that clusters copper-circuit addresses with DBSCAN and coordinates six specialist agents into prioritized build plans, cutting manual fiber build planning effort by 65%.
  • Built a LangGraph routing agent that triages 278,350 tickets at 90% accuracy using vendor-specific, glossary-grounded LLM prompts — eliminating ~47,000 hours of annual L1/L2 work for a projected $2.46M in cost avoidance.

MLOps Engineer (Contract)

Apple Inc.·Data Platform Efficiency
Apr 2025 – Aug 2025Dallas, TX

Software & AI Engineer

Eminent Services Corporation
May 2024 – Apr 2025Frederick, MD

ML Research Assistant

University at Buffalo
Jan 2024 – May 2024Buffalo, NY

Solutions Engineer

Swym Corporation
Aug 2022 – Aug 2023Bangalore, India
Products

Things I've shipped.

prism-mem

LiveSolo build

Post-session memory for AI coding agents.

An open-source CLI tool I built solo. After every git commit, it reads Claude Code session transcripts and git diffs, extracts structured knowledge as (subject, predicate, object) triples, links them by cosine similarity, auto-detects stale facts, and rewrites CLAUDE.md, .cursorrules, and AGENTS.md. In a 3-session demo, it compressed 411,463 bytes of raw transcripts into 11,707 characters of structured knowledge — a 35x reduction.

kg-gensentence-transformersnumpySQLiteFastMCPFastAPILiteLLMPython
PyPIGitHubLearn moreRead story

What it does

  • Extracts (subject, predicate, object) triples from Claude Code session transcripts and git diffs via kg-gen and an LLM
  • Links related triples by cosine similarity and marks contradicted facts stale automatically — no manual input
  • Rewrites CLAUDE.md, .cursorrules, and AGENTS.md after every git commit via a post-commit hook
  • MCP server exposes get_context, query_knowledge, and crystallize tools to Claude Code, Cursor, and Codex

ApplyAI

LiveSolo build

Apply to jobs 10× faster.

A full-stack job application assistant I built solo — Chrome extension, web dashboard, and AI backend. It autofills application forms using a LangGraph agent, tracks every application in one place, and scores your resume against each job posting.

LangGraphGeminiFastAPINext.jsChrome Extension MV3Supabase
Live AppChrome StoreGitHubRead story

What it does

  • LangGraph DAG agent autofills entire forms — dropdowns, React Select, and file uploads included
  • Resume-to-job match scoring with matched and missing keyword breakdown
  • Application tracking dashboard with status pipeline, KPI cards, and charts
  • Job board discovery across Lever, Ashby, and Greenhouse via SERP search
Projects

Things I've built.

apply-ai-extension

Full-stack job application assistant with a Chrome extension, LangGraph autofill agent, resume match scoring, and application tracking dashboard. Built solo and live on the Chrome Web Store.

LangGraphGeminiFastAPINext.js+2
GitHubLive

springcommerce

7

Production-grade microservices e-commerce platform built with Spring Boot, secured with Keycloak, orchestrated on Kubernetes, and monitored via Grafana + Prometheus.

Spring BootKubernetesGrafanaKeycloak+1
GitHub

secure-stack

End-to-end DevSecOps pipeline implementing a three-tier architecture on AWS EKS with ArgoCD, GitOps workflows, and automated security scanning.

AWS EKSArgoCDGitOpsDevSecOps+1
GitHub

atmo-flow

Real-time weather and air quality data pipeline on GCP using Cloud Composer (Airflow), PySpark transformations, and interactive visualizations.

GCPAirflowPySparkData Pipeline
GitHub
View all projects
Writing

Things I've written.

May 31, 2026

prism-mem: Automatic Knowledge Extraction for AI Coding Agents

Built prism-mem, an open-source CLI that reads Claude Code session transcripts and git diffs after every commit, extracts (subject, predicate, object) triples via kg-gen, links them by cosine similarity, auto-detects stale facts, and rewrites CLAUDE.md automatically — no manual context updates needed.

AI AgentsKnowledge GraphPython
Read on dev.to
Mar 6, 2025

I Got Tired of Filling Out the Same Form 50 Times, So I Built an AI to Do It

Built ApplyAI, a Chrome extension that automates job application form filling using a LangGraph agent and Gemini AI, cutting the process from 10 minutes to under 10 seconds.

AI AgentsAutomationPython
Read on dev.to
Jan 20, 2025

AtmoFlow: Real-Time Weather and Air Quality Insights

End-to-end data pipeline on GCP using Cloud Functions, Pub/Sub, Dataproc, and BigQuery to ingest, process, and visualize environmental data with Looker dashboards.

Data EngineeringGoogle CloudPySpark
Read on dev.to
Mar 12, 2025

Spring Commerce: Building a Resilient E-commerce System with Spring Boot Microservices

A complete microservices e-commerce system with four core services communicating via REST and Kafka, secured with Keycloak, and monitored through the Grafana observability stack.

Spring BootMicroservicesKubernetes
Read on dev.to
Mar 12, 2025

Spring Core Fundamentals: A Beginner Guide

A practical guide to Spring's core concepts: dependency injection, inversion of control, bean management, and aspect-oriented programming for building maintainable Java apps.

JavaSpring BootBackend
Read on dev.to
All posts on dev.to
Skills

The toolkit.

Years of building at the frontier, from LLM fine-tuning to distributed data pipelines to cloud-native deployments.

Agentic & ML
LangGraphLangChainCrewAIAutogenPyTorchTensorFlowGoogle ADKMCPA2A
RAG & Knowledge
Hybrid RAGLlamaIndexNeo4jElastic Vector DBpgvectorCohere RerankSemantic Search
MLOps & LLMOps
RayONNXMLflowNVIDIA TritonLangSmithLangFuseKubeflowW&BAxolotl
Data Engineering
SparkAirflowKafkadbtSnowflakeTimescaleDBETL/ELT
Cloud & DevOps
AWSGCPAzureKubernetesTerraformDockerArgoCDCI/CD
Programming
PythonTypeScriptJavaSQLFastAPISpring BootNode.jsReactNext.js
Education

Where it started.

MS, Computer Science

University at Buffalo

Buffalo, NY
GPA 3.8 / 4.0

BTech, Computer Science

Vellore Institute of Technology

Chennai, India
GPA 3.9 / 4.0
Contact

Let's work
together.

I'm open to full-time roles and interesting collaborations. If you're building something ambitious in the AI space, I want to hear about it.

Or just chat with my AI. It knows everything about me.

© 2026 Rahul Reddy Talatala.

United States