AI Engineer & Full-Stack Developer from Chennai, India, with production experience shipping LLM-powered applications using the Claude API, Gemini API, and open-source models. I've built and deployed platforms end-to-end — from prompt engineering and multi-modal AI pipelines to React frontends and Dockerized cloud deployments.

• 🎓 B.Tech in Information Technology (Honors) — Anna University, CGPA 8.51/10
• 🚀 Shipped two live platforms: Elevra (AI career suite) and ScopeAI (scope creep detection)
• 🤖 Focused on LLM orchestration, RAG, and fine-tuning (LoRA/QLoRA/PEFT)
• 📬 Bharathsiva453@gmail.com

Full-Stack Developer Intern — Jasmin InfoTech Pvt. Ltd. · Jun 2026 – Jul 2026 Led a 3-member team to build and ship ScopeAI from zero to live in 4 weeks. Designed a hybrid AI architecture combining Gemini API feature-level analysis with a deterministic capacity engine for explainable timeline and risk output.

Full-Stack Development Intern — Approtech R&D Solutions · Jun 2025 – Aug 2025 Re-engineered the authentication module (Flask + JWT), cutting API response time by 40% through query optimization and connection pooling. Designed and documented 12+ RESTful endpoints for independent frontend delivery.

Production-deployed full-stack career suite — my most comprehensive project

An all-in-one career platform built for freshers to land jobs faster. Combines Resume Intelligence, Mock Interview Coaching, Job Description Gap Analysis, Live Job Search, a Career Readiness Index, AI Career Roadmaps, and a Chrome Extension for one-click job applications.

🔗 Live App: elevraa.me  |  Stack: React 19 · TypeScript · FastAPI · PostgreSQL · Claude API + Gemini API (dual-LLM hot-swap) · Docker · Azure

• Engineered a dual-LLM orchestration layer with automatic provider hot-swap for zero-downtime model switching
• Built an AI Mock Interview Coach running 3 simultaneous ML streams: faster-whisper transcription, DeepFace sentiment, MediaPipe eye-contact tracking
• Shipped 7 modules: ATS resume scoring, JD gap analysis, live job search, Career Readiness Index, AI roadmap, and a Chrome MV3 extension

Catches scope creep before it derails a project

Detects scope creep in real time, runs what-if scenario modeling (team size, deadlines, feature changes), and provides AI-backed capacity analysis with PDF/CSV export. Built with a hybrid AI architecture — Google Gemini handles feature-level analysis while a deterministic engine computes project-level metrics.

🔗 Live Demo: scope-shift.vercel.app  |  Stack: React 19 · TypeScript · Express.js · Prisma ORM · MySQL · Gemini API · Tailwind CSS v4

Final Year Capstone Project

Analyzes resumes for ATS compatibility using a weighted scoring algorithm (Skills 40% · Experience 30% · Education 20% · Format 10%), extracts skills with spaCy NLP, and generates improvement suggestions via Google Gemini. Includes a full resume builder with 5 professional templates and live PDF preview.

🔗 Live Demo: resume-ai-frontend-28cb.onrender.com  |  Stack: FastAPI · React.js · SQLite · Google Gemini API · spaCy · PyJWT · Tailwind CSS

AI-driven interview practice with real-time speech and emotion analysis

Conducts AI-driven mock interviews with live webcam recording (WebRTC). Evaluates speech via Whisper (transcription, filler words, pace), tracks eye contact and emotions via MediaPipe + DeepFace, and grades answer quality through Claude API / GPT-4. Delivers detailed performance reports with progress tracking.

Stack: React 18 · TypeScript · FastAPI · MongoDB · AWS S3 · Redis · Whisper · DeepFace · Claude API

Custom fine-tuned model for dev/infra troubleshooting

A Mistral-7B model fine-tuned with LoRA adapters (Rank 16, Alpha 32) on curated troubleshooting instruction–response pairs. Resolves issues across Docker, Git, Linux, CI/CD, backend APIs, databases, and cloud deployments. Deployed via a lightweight Gradio web interface.

Stack: Python · Mistral-7B-Instruct · LoRA / PEFT · Hugging Face · Gradio · Transformers

"Building applications that combine smart design with intelligent AI — one commit at a time."