I'm a full-stack engineer and AI/ML builder based in Bengaluru, currently pursuing my B.Tech in Computer Science at PES University. My work sits at the intersection of generative AI, cloud-native architecture, and product engineering — I design systems end-to-end, from data pipelines and model integration through to the interfaces people actually use.
I placed 1st nationally at the Google Gen AI Exchange Hackathon (Google Cloud × Hack2Skill), competing against 270,000+ developers, and hold a second 1st-place finish at Inferentia 2.0. My current focus is shipping production-grade GenAI products on Google Cloud — combining Vertex AI, Gemini, and Search Grounding with scalable Next.js/FastAPI architectures.
I care about systems that hold up outside a hackathon judging room — sensible architecture, real cost tradeoffs, and code that's still readable six months later.
- 🎯 Open to: SDE, AI/ML Engineer, and Full-Stack internship roles
- 🛠️ Currently building: a production GenAI platform on GCP (FastAPI + Next.js + Vertex AI)
- 🌱 Currently exploring: agentic AI workflows, multi-modal systems, system design fundamentals
- 💬 Ask me about: GenAI on GCP, Vertex AI / Gemini integration, full-stack architecture, building under hackathon time pressure
Languages
Frontend
Backend & Databases
Cloud, DevOps & Tooling
IoT & Embedded
| Domain | Proficiency | Details |
|---|---|---|
| Generative AI (Gemini + Vertex AI) | Advanced | Shipped two production GenAI apps using Gemini APIs with Search Grounding; routed Vertex AI workloads across regions for grounding availability |
| Pipeline Architecture | Advanced | Migrated a multi-agent LangGraph scraping pipeline to a leaner Search-Grounded Gemini approach — cut Docker image size and simplified the codebase significantly |
| ML Model Building (scikit-learn) | Intermediate | Built price-forecasting and disease-detection models using BigQuery ML for real-time agricultural data |
| Computer Vision (Vision API) | Intermediate | Integrated Google Vision API for real-time crop disease detection |
| Cloud AI Infrastructure | Intermediate | Deployed AI workloads on Cloud Run with Firestore-backed caching layers |
🥇 CareerLens — Personalized AI Career & Skills Advisor
GenAI-powered career guidance platform built for the Google Cloud Gen AI Exchange Hackathon. Provides resume scoring, job matching, skill-gap analysis, and personalized learning paths using Vertex AI and Gemini — evaluated against 270,000+ developers nationwide and selected as the overall winner.
| Aspect | Details |
|---|---|
| Stack | Next.js, React.js, Node.js, Firebase, Firestore, Vertex AI, Gemini APIs, Firebase Auth |
| Scale | Evaluated by Google Cloud among 270,000+ developers and thousands of teams nationwide |
| Performance | Advanced through three competitive rounds — Top 100 → Top 10 → Winner |
| Security | Firebase Authentication with scoped Firestore security rules |
| Impact | 🥇 1st Place — Google Gen AI Exchange Hackathon · ₹2,00,000 prize |
| Repository | github.com/BharathPESU/CareerLens |
CareerLens was architected to go beyond a single LLM call — resume scoring, job matching, and skill-gap analysis are handled as distinct stages, with Gemini orchestrating personalized output across each. The win came down to treating it like a real product: clean auth, structured data in Firestore, and a UI built to be demoed live under judging pressure.
🥇 AgriSense — AI-Powered Smart Agriculture Platform
An IoT and AI-based platform for real-time monitoring of soil moisture, temperature, crop health, and environmental conditions — combining sensor hardware with cloud ML for disease detection, irrigation automation, and price forecasting.
| Aspect | Details |
|---|---|
| Stack | Python (Flask), React.js, TypeScript, scikit-learn, GCP (BigQuery ML, Vision API, Earth Engine, Firebase), Raspberry Pi, ESP32, MQTT |
| Scale | Real-time multi-sensor ingestion pipeline feeding a live ML inference layer |
| Performance | Designed, built, and demoed within a 24-hour hackathon window |
| Security | Firebase-backed data layer with cloud-native access control |
| Impact | 🥇 1st Place — Inferentia 2.0 Hackathon · ₹75,000 prize |
| Repository | github.com/BharathPESU/hackthon1 |
The hardest part of AgriSense wasn't the ML — it was getting Raspberry Pi and ESP32 sensors to report reliably over MQTT into a pipeline that BigQuery ML and Vision API could actually consume in real time, end to end, in a single day.
🔗 PeerLink — Offline Peer-to-Peer File Sharing & Messaging
An Android application enabling fully offline peer-to-peer file sharing and real-time messaging over local Wi-Fi — zero internet required. Peers are auto-discovered via UDP broadcast, with file transfer handled over TCP using a multi-peer chunk-download strategy for speed.
| Aspect | Details |
|---|---|
| Stack | Kotlin, Android SDK, TCP/UDP Sockets, Kotlin Coroutines, SharedFlow, ViewModel, RecyclerView, Material UI |
| Scale | Fully offline, multi-peer chunked-transfer architecture |
| Performance | Parallel multi-peer chunk downloads for faster transfer speeds |
| Security | Local-network only — zero internet dependency, no external data exposure |
| Impact | Demonstrates systems-level networking design beyond typical web/app projects |
| Repository | github.com/BharathPESU/PeerLink |
PeerLink was a deliberate move away from cloud-dependent apps — the interesting engineering problem was peer discovery and chunked transfer reliability on flaky local Wi-Fi, not the UI.
🏅 META — AI Hackathon Project (OpenEnv × Meta × Scaler)
An AI-powered project built for the OpenEnv Hackathon, hosted by Meta and Scaler. Competing as team Error_404_Not_Found, the project was selected among the Top 800 teams from over 31,000 national registrations.
| Aspect | Details |
|---|---|
| Stack | Python, AI/ML, React |
| Scale | Built for a hackathon with 31,000+ national registrations |
| Performance | Selected into the Top 800 teams |
| Security | Not applicable — early-stage hackathon prototype |
| Impact | Finalist, HackArena 2.0 · Mozilla.ai Track · Bangalore Zonals |
| Repository | github.com/BharathPESU/META |
🎓 KCET College Predictor — GenAI-Powered Admissions Platform
A full-stack platform that helps students predict realistic college and branch options based on KCET rank, using live search-grounded AI rather than static cutoff tables. Deployed end-to-end on Google Cloud.
| Aspect | Details |
|---|---|
| Stack | FastAPI, Next.js 14, Tailwind CSS, shadcn/ui, Vertex AI (Gemini + Search Grounding), Firestore, Cloud Run, Firebase Hosting |
| Scale | Production GCP deployment with dedicated Firestore collections for cutoffs and feedback caching |
| Performance | Re-architected from a multi-agent LangGraph scraping pipeline to a Search-Grounded Gemini 2.0 Flash approach — dramatically smaller Docker images, simpler codebase, more reliable output |
| Security | Cloud Run with region-aware routing (asia-south1 / us-central1) for Search Grounding availability |
| Impact | End-to-end AI product — data pipeline design through production deployment with CI/CD via Cloud Build |
| Repository | Private — available on request |
This project's real lesson was knowing when to simplify: the original scraping-based multi-agent pipeline looked impressive on paper but was unreliable in practice. Replacing it with Gemini + Search Grounding traded some theoretical flexibility for something that actually worked in production.
Technical Lead — Team Error_404_Not_Found
PES University & National Hackathon Circuit · 2024 – Present
Leading a close-knit team of developers through end-to-end builds for national-level AI/ML hackathons — from architecture decisions to live demo, typically under 24–48 hour constraints.
- Architected and shipped CareerLens, a GenAI career advisory platform, from concept to a 1st-place finish among 270,000+ developers
- Directed technical execution for AgriSense, an IoT + AI agriculture platform, to a 1st-place finish at Inferentia 2.0
- Led the team's submission to the Meta × Scaler OpenEnv Hackathon, advancing to the Top 800 of 31,000+ registrations
- Made build-vs-buy architecture calls under time pressure — including a mid-build pivot from a LangGraph scraping pipeline to a Search-Grounded Gemini approach to hit reliability targets
Team Leadership System Architecture Rapid Prototyping GenAI Integration Cloud Deployment
| Recognition | Details |
|---|---|
| 🥇 1st Place — Google Gen AI Exchange Hackathon | Google Cloud × Hack2Skill · 270,000+ developers nationwide · Top 100 → Top 10 → Winner · ₹2,00,000 prize · Built CareerLens |
| 🥇 1st Place — Inferentia 2.0 Hackathon | PES University · State-level, 24-hour build · ₹75,000 prize · Built AgriSense |
| 🏅 Top 100 Finalist — Google Gen AI Exchange | Shortlisted from 90,000+ participants nationwide |
| 🏅 OpenEnv Hackathon Finalist | Meta × Scaler · Top 800 of 31,000+ registrations · HackArena 2.0, Mozilla.ai Track · Team Error_404_Not_Found |
☁️ Google Cloud — Skill Badges
current_focus:
learning:
- Advanced system design patterns
- Agentic AI orchestration & multi-agent workflows
- Competitive programming depth (DSA)
building:
- Production GenAI platform on GCP (FastAPI + Next.js + Vertex AI)
- Personal portfolio v2 — AI Engineer rebrand
exploring:
- Multi-modal AI systems
- RAG architectures beyond Search Grounding
open_to:
- SDE internships
- AI/ML Engineer internships
- Full-stack engineering roles
- Open-source collaboration
Building things that work past the demo — one production system at a time.