██╗ ██╗ █████╗ ██████╗ ██╗
██║ ██║██╔══██╗██╔══██╗██║
███████║███████║██████╔╝██║
██╔══██║██╔══██║██╔══██╗██║
██║ ██║██║ ██║██║ ██║██║
╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝
name : HARINARAYANAN U
location : Kerala, India
education : B.Tech CSE @ Amrita Vishwa Vidyapeetham (2026)
focus : AI-powered backends · LLM systems · iOS · production ML
status : Open to roles in AI/ML engineering & backend systems
publications: 2× Scopus-indexed (ICMLDE 2025, I4Tech 2025)Currently building systems where language models meet real-world data — from Text-to-SQL pipelines to multimodal emotion engines.
RevGain AI — Software Developer Intern (Sep 2025 – Feb 2026)
► Built Text-to-SQL microservice (FastAPI + PostgreSQL)
→ schema ingestion, dynamic filtering, SQL validation
► ~85% execution accuracy on complex queries via few-shot prompting
→ ~40% reduction in hallucinated SQL vs. zero-shot baseline
► <600ms p95 API latency with Dockerized, read-only credential setup
TiVo — iOS Developer Intern (Aug 2024 – Nov 2024)
► Refactored AVPlayer caching + prefetch logic
→ ~25% reduction in buffering incidents
► Redesigned 8+ SwiftUI components
► 50+ collaborative code reviews
🎵 Sonar — Emotion-Aware Music Recommender React · FastAPI · PostgreSQL · Transformers · Docker
- Confidence-weighted text + speech emotion fusion → personalized Spotify playlists
- Hierarchical emotion classification: GoEmotions (text) + WavLM embeddings (speech)
- SHAP-based explainability for transparent predictions
- Adaptive personalization via contextual bandit (skip/like/listen signals)
🍣 SushiGo — Full-Stack Food Delivery App Swift · Django REST · PostgreSQL · GPS
- Full iOS client (Swift) + Django REST backend with auth, cart, checkout, order lifecycle
- Role-based workflows: customer / driver / admin
- Stripe payment integration + backend verification
- Real-time order tracking via background GPS + push notifications
@inproceedings{NeuroCode2025,
title = {NeuroCode: Dual-Model LLM System for Proactive Bug Prediction},
venue = {ICMLDE 2025 (Scopus)},
highlight = {LLM + CodeBERT → 89% accuracy on 10K+ code samples}
}
@inproceedings{PrivMod2025,
title = {PrivMod: Federated Learning for Privacy-Preserving Content Moderation},
venue = {I4Tech 2025 (Scopus)},
highlight = {Federated CNN-BERT hybrid → 94% accuracy, ε=0.5 differential privacy}
}Languages
Backend & APIs
AI / ML
iOS
Cloud & Tools
currently_building = ["AI-powered backend systems", "LLM-based applications"]
looking_to_collab = ["real-world ML projects", "backend systems", "research AI"]
currently_learning = ["LLM optimization", "scalable backend design", "production ML research", "SwiftUI advanced patterns", "Core ML on-device inference"]
ask_me_about = ["Text-to-SQL", "FastAPI", "iOS / SwiftUI", "end-to-end AI products"]