hemanth.s
AI/ML Engineer · Bengaluru, India

I build productionGenAI systems.

End-to-end — from client brief to live deployment — across AWS, Azure & GCP.

See the workTry the cost lab
15+
production AI systems shipped
3
clouds — AWS, Azure, GCP
35%
LLM cost cut, one client
20%
latency cut, same system

$ cd ~/work

Five systems, three clouds.

Each one built and shipped to production — real users, real traffic, real constraints.

Flagship project

LiveAWSEdTech

EdTech Learning Platform

30,000+ users / month

Production platform on AWS Bedrock (Claude 4 Sonnet) powering personalized note generation, an AI job-search assistant, voice-based interview assessment, multi-template resume generation, personalized roadmaps, and a course-specific RAG chatbot.

›Personalized note generation
›AI job-search assistant
›Voice interview assessment
›Multi-template resume builder
›Personalized learning roadmaps
›Course-specific RAG chatbot
AWS BedrockClaude 4 SonnetRAGVoice AI

→ Gave the client a single, personalized, end-to-end learning experience that moved the needle on student engagement and career readiness.

Selected work

LiveAWSBanking / Fintech

Multilingual Post-Call Analytics

6,000+ calls / month · 90%+ accuracy

Analytics pipeline for a major Indian bank using Transcribe, Comprehend, Bedrock, and CloudFront to score sentiment, flag fraud signals, and detect call-resolution status — backed by a live analytics dashboard.

TranscribeComprehendBedrockCloudFront

→ Let the client catch fraud risk and unresolved issues faster, cutting manual call-review effort.

LiveAzureRetail Enterprise

WhatsApp AI Agent for Retail

24/7 self-service

Azure AI Foundry agent orchestrating RAG product search, NL-to-SQL, user verification, and CRM-integrated ticket creation over WhatsApp.

Azure AI FoundryRAGNL-to-SQLCRM

→ Round-the-clock customer self-service that eased load on the client's support team.

LiveAWSEsports

Recommendation & Content Moderation

Real-time stream moderation

Amazon Personalize generating game recommendations from engagement and purchase data, paired with Amazon Rekognition moderating live-stream video in real time to flag abusive, NSFW, or spam content.

Amazon PersonalizeRekognitionLive Video

→ Boosted engagement through relevant recommendations while keeping the platform safe and trusted.

LiveGCPRetail Enterprise

RAG Chatbot for Retail Website

Low-latency, on-site support

Conversational RAG assistant on Vertex AI built for scale, cost-efficiency, and low-latency response on a retail platform's own website.

Vertex AIRAGCost optimization

→ Instant, accurate on-site answers, reducing dependency on manual support channels.

$ run cost-optimizer --interactive

This is how the 35% happened.

Drag the sliders back to zero to see what the same system costs and takes without model routing and semantic caching turned on. Defaults reflect what actually shipped.

100%

Simple queries route to a smaller model; complex analytical calls go to the larger one.

100%

Repeated and near-duplicate queries are served from cache instead of hitting the model again.

cost / 1k queries$5.46

baseline $8.40

p50 latency1920ms

baseline 2400ms

monthly · 100k queries

$546 / $840

cost reduction

35%

real design: redis semantic cache on sentence-embeddings + lighter model for lookups, larger model reserved for complex analytical calls with dry-run validation.

$ cd ~/experience

Where I've built.

Jan 2025 — Present

Associate AI/ML Engineer

Rapyder Cloud Solution

  • Primary engineer on 15+ AI/ML solutions across AWS, Azure, and GCP, including multiple production-grade deployments — owning client conversations end-to-end.
  • Delivered GenAI systems across Manufacturing, Healthcare, E-commerce, Gaming, and Fintech, matching architecture (RAG, multi-agent, voice/chat) to each use case.
  • Cut LLM API costs 35% via model right-sizing, prompt optimization, and response caching.
  • Cut system latency 20% via parallelized pipelines, optimized DB read/write patterns, and targeted caching.

Education

B.E., Computer Science

ACS College of Engineering, Bengaluru — CGPA 7.57, 2024

// about

Engineered, not assembled.

AI/ML Engineer with hands-on experience delivering 15+ production-grade GenAI solutions across AWS, Azure, and GCP. Primary engineer on every build — owning the path from client requirements through deployment — with a track record of cutting LLM inference costs and system latency through deliberate architecture choices, not just model swaps.

Currently at

Rapyder Cloud Solution

Associate AI/ML Engineer

Jan 2025 — Present

How I work

  1. 01

    Understand the ask

    Sit with the client's requirement directly, not a handed-down spec — scope the real constraint before any architecture gets drawn.

  2. 02

    Design for the use case

    RAG, multi-agent, voice, or plain NL-to-SQL — the pattern follows the problem, not the other way round.

  3. 03

    Build & ship

    Python services, cloud-native infra, and model integrations, taken to a live, production-grade deployment.

  4. 04

    Optimize in production

    Model routing, caching, and pipeline tuning after launch — where the real cost and latency wins happen.

// capabilities

The toolkit.

Languages

PythonGit

Cloud

AWSAzureGCPBedrockLambdaRekognitionSageMakerECSEC2Azure AI FoundryAzure FunctionsAzure AI SearchAzure OpenAI

GenAI & ML

LangGraphLangChainLlamaIndexHuggingFacePyTorchTensorFlowRAGMulti-Agent SystemsPrompt EngineeringFine-tuningMLOpsNLP-to-SQL

Data & Infra

Aurora pgvectorPineconeFAISSTerraformCDK

// certifications

Verified, not just claimed.

ML
Amazon Web Services

AWS Certified Machine Learning – Associate

Verify ↗
DB
Databricks

Databricks Certified Generative AI Engineer Associate

Verify ↗

// now

$ status = "Building production GenAI systems at Rapyder Cloud — owning everything from client brief to live deployment across AWS, Azure & GCP."▋

Last updated · Aug 2025 · Bengaluru, India

// built with

Next.js 14· App Router
TypeScript· Strict mode
Tailwind CSS· Custom design tokens
GSAP· ScrollTrigger animations
Lenis· Smooth scroll
JetBrains Mono· Monospace font
Inter· Body font
Vercel· Deployment

// contact

Have a system that needs building — or an expensive one that needs fixing?

Based in Bengaluru, India. Open to full-time roles and select production GenAI engagements.

hemanths01234@gmail.comLinkedIn ↗
hemanths01234@gmail.comlinkedin