About Me
Who I Am
Yelisetty Yaswanth Sai
AI & Cloud Engineer
Core Specializations
"Building intelligent, scalable infrastructure that bridges cutting-edge AI with real-world applications. Focused on deploying production-ready solutions that solve complex challenges."
Production-Ready Expertise
I specialize in building intelligent systems that bridge the gap between cutting-edge AI and real-world applications. From architecting RAG pipelines to deploying scalable cloud infrastructure, I turn complex technical challenges into elegant, production-ready solutions.
Journey
Education
B.Tech Computer Science & Engineering
Dhanekula Institute of Engineering and Technology
Currently maintaining 7.96 CGPA (5th semester). Specialized in AI/ML, Cloud Computing, and Full-Stack Development. Building real-world projects with Google Gemini AI and Azure cloud infrastructure.
Intermediate (MPC)
Sri Dhanalakshmi Junior College
Completed intermediate education with strong foundation in Mathematics, Physics, and Chemistry, scoring 89%. Built early interest in programming and problem-solving.
Secondary Education (SSC)
Alfried Nobel High School
Completed secondary education with 92%, establishing fundamentals in analytical thinking and academic excellence.
Selected Work
Things I've built.
CloudScale-RL
AI-Powered Cloud Auto-Scaling Environment
Machine Learning / DevOps • April 2026 • Completed
A high-fidelity Digital Twin reinforcement learning environment for cloud auto-scaling, built for the Meta × Hugging Face OpenEnv Hackathon 2026. Solves over-provisioning waste and under-provisioning outages through learned autonomous scaling policies.
3
Tasks
16
Observation Fields
~2000
Lines of Code
3/3 Passed
Validation Checks
Before vs After
Before (Rule-Based)
❌ Reactive scaling
❌ Fixed thresholds
❌ No boot delay model
❌ Crashes on spikes
After (CloudScale-RL)
✅ Predictive scaling
✅ Learned policies
✅ 2-step warm-up
✅ Survives 10x load
Highlights
- • Digital Twin simulation with server warm-up delays
- • Exponential latency physics (quartic transfer function)
- • Deployed to Hugging Face Spaces with Docker
- • LLM inference with fallback heuristics
Exponential Latency Physics
def _calculate_latency(self, rps: float, servers: int) -> float:
load_ratio = rps / (servers * CAPACITY_PER_SERVER)
latency = BASE_LATENCY + (load_ratio ** 4) * 100 # The Wall
return min(latency, CRASH_THRESHOLD + 100)UniPeasy
Unified Student Success Platform
EdTech / AI Product • Feb 2026 Launch • Live
Founder & Technical Lead • Timeline: 6 Months
From idea to impact, UniPeasy went from a student pain-point in May 2025 to a live AI-powered success platform in Feb 2026. It solves fragmented resources, exam stress, skill gaps, and trust issues with verified materials and authenticated opportunities. Today, 150+ students actively use UniPeasy.
The Wall of Challenges
- • Complexity Overload — dense information leads to passive reading
- • Stressful Exam Prep — no strategy, last-minute cramming
- • The Skill Gap — academics ignore real-world skills
- • Resource Fragmentation — hours wasted on unverified sources
- • Knowledge Loss — no systematic revision loop
Our Intelligent Solutions
- • AI Learning Assistant — simple explanations, analogies, mind maps, adaptive quizzes
- • AI Exam Strategist — personalized Pomodoro-based timetables
- • AI Skill Accelerator — structured tracks with instant AI feedback
- • Centralized Materials Hub — topper-verified by branch, year, subject
- • Memory Palace — save AI insights for spaced repetition
- • Authenticated Internships — manually verified opportunities only
Timeline
May 2025 → Feb 2026
6-month build from idea to live deployment
Leadership
Founder & Technical Lead
Led a team of 9 across full product lifecycle
Core Stack
Next.js 14 · Firebase · Gemini AI
Full-stack architecture, UX strategy, deployment
Adoption
150+ Students Actively Using UniPeasy
Built from a real student problem and now delivering daily impact through verified resources, AI learning workflows, and trusted opportunities.
Trust & Safety Commitment
- • No fake internships or hackathons listings
- • Community-powered suggestions with manual verification
- • No data misuse or third-party selling
- • Built by students, protecting student interests
AutoTask
AI-Powered Task Management via WhatsApp
AI Productivity / Cloud Backend • v2 Production • Live
Product Builder & Backend Architect • Timeline: Ongoing Evolution
AutoTask started from a personal productivity problem: forgetting important tasks. I transformed a simple WhatsApp reminder idea into a production-ready AI task platform with robust recurrence logic, strict parsing guardrails, and reliable cloud deployment.
Personal Problem → Product
I built AutoTask to solve my own daily problem: I frequently forgot important tasks. Since I use WhatsApp all the time, I designed a natural-language reminder flow where I can send messages like "remind me about box at 9am" and receive an exact reminder at the right time. This became a production-grade assistant that genuinely improved my day-to-day execution.
Hey! 👋
Production Architecture (Live v2)
- • Frontend: Vercel-hosted web dashboard
- • Backend: Node.js + Express controllers/services
- • Database: MongoDB API on Azure Cosmos DB
- • Messaging: Twilio WhatsApp webhook integration
- • Runtime: Azure VM + PM2 + Nginx reverse proxy
Core Capabilities (Live)
- • Natural-language task extraction with Gemini parser
- • Timezone-aware reminders and advanced recurrence engine
- • Exception handling: skip weekends/holidays, pause until, custom dates
- • WhatsApp 'done' command marks latest pending reminder complete
- • Structured JSON extraction and normalization guardrails
Status
Production (v2 Active)
Stable with monitored health checks
Validation
Smoke + End-to-End Tested
Parser, recurrence, scheduler, production endpoints
Known Transient Issue
Gemini 429 on burst tests
Appears under rapid sequential test bursts
Next Phase: RAG Integration (In Progress)
- • Vector store: ChromaDB local persistent path
- • Embeddings: text-embedding-004
- • Retrieval: top-K similar completed tasks for context grounding
- • Generation: context-injected prompts for parsing/planning/Q&A
- • Backfill: one-time embedding of historical completed tasks
Expertise
Technology Stack
A comprehensive toolkit spanning AI, cloud infrastructure, and modern web development.
Artificial Intelligence
Cloud Infrastructure
Application Development
Get in Touch
Contact
yaswanthsaiyelisetty@gmail.comOpen to AI, Cloud, and Full-Stack opportunities. Let's build dependable systems.
Hi! I'm Yaswanth's AI assistant. Ask me anything about his experience, projects, or skills.