Turning models into systems that hold up in production.
I'm a final-year Data Science student at CHRIST University who'd rather ship something than just prototype it. Recent work: fine-tuning MuRIL for Hinglish sentiment analysis, building CareerMatch (a resume-screening platform on Llama 3), and launching Mana Products, a production e-commerce site on Next.js and Supabase. Currently looking for a Data Science / ML Engineering internship.
Selected Work — 05 Projects
-
01 · 2026
Mana Products
Full-Stack E-Commerce Catalog & Admin Platform
Next.js 15 React 19 TypeScript Supabase Tailwind CSS VitestProduction e-commerce catalog and admin platform (manaproducts.in) for a home-based spice and pickle business. Custom session auth (bcrypt + HMAC-SHA256 via Web Crypto API), Postgres Row-Level-Security with all writes routed through server-only Server Actions, an admin CRUD dashboard with image upload and unit tests, and WhatsApp-based deep-link ordering in place of a payment gateway.
-
02 · 2026
Hinglish Sentiment Analyzer
Code-Mixed NLP with Fine-Tuned MuRIL
Python PyTorch Hugging Face Transformers Google MuRIL StreamlitFine-tuned
google/muril-base-casedfor three-class sentiment classification on code-mixed Hindi-English text, where English-trained models degrade sharply on Romanized Hindi. Trained on ~17,000 cleaned samples, reaching 64.6% accuracy and 0.647 weighted F1 on a held-out ~1,700-sample set — nearly double the random-chance baseline. Streamlit interface supports single-text and batch CSV prediction with confidence bands. -
03 · 2026
CareerMatch AI
AI Resume-Screening & Job-Matching Platform
Python Streamlit Ollama / Llama 3 Sentence Transformers Scikit-learnParses a candidate's resume, scores it against a target job description with embedding-similarity job-fit matching, and flags ATS-compatibility and skill gaps. A Llama 3 career-assistant chatbot, served locally via Ollama, answers resume-feedback and career-guidance questions grounded in the parsed profile. Built during a Data Science & ML internship at IPSR Solutions; tested with real users.
-
04 · 2026
Axiom
Token-First Design System & Governance Framework
TypeScript Style Dictionary Tailwind CSS JSON ADRsA token-first, multi-layer design system: JSON design tokens compiled by a Style Dictionary pipeline into typed CSS variables, a semantic theme layer, and components — enforcing a strict one-directional dependency chain. Backed by a ~12-section governance spec and an
AI_CONTEXT.mdfor AI coding agents, then validated by using it to build this portfolio. -
05 · 2025–26
Mechanical Department Monitoring Platform
Web Development, Data Analytics
Web Development Data AnalyticsA web platform tracking machine utilization for CHRIST University's Mechanical Department, with dashboards surfacing usage patterns to support predictive maintenance planning. Owned deployment and testing end-to-end; in active use by the department.
Internal deployment — not public
Studying data science, building things that ship.
I'm a final-year CSE (Data Science) student with hands-on experience across applied NLP/transformer fine-tuning and full-stack product engineering — from training a MuRIL classifier on code-mixed Hinglish text to shipping a production e-commerce platform on Next.js and Supabase. I'm most interested in the gap between a notebook that works and a system that stays working in front of real users.
Experience
Data Science & Machine Learning Intern · Kochi
Built CareerMatch end-to-end — resume parsing, semantic job-fit scoring, ATS/skill-gap analysis, and a Llama 3 career-assistant chatbot served locally via Ollama.
Software Development Intern · Bengaluru
Built a Sudoku game with backtracking-based puzzle generation and solution validation; tested and optimized solve-validation logic across difficulty levels.
Education
B.Tech, Computer Science Engineering (Data Science) · Minor in General Management · Bengaluru · CGPA 8.01/10.0
Skills
Let's talk.
Open to Data Science / ML Engineering internships and collaboration. The fastest way to reach me is email.
nived@nivedkrishnan.com