Production-first engineering across cloud ML, autonomous workflows, and resilient backend systems.
About / Life Path
I build software where AI is part of the operating layer, not a bolt-on feature. That means autonomous workflows, dependable APIs, and backend systems designed for messy real-world traffic. I am currently pursuing an M.Sc. in Artificial Intelligence & Machine Learning at Coimbatore Institute of Technology, while spending most of my time building and shipping. In industry work, I have built FastAPI services, AWS data and ML pipelines, and risk modeling flows that moved from experiments into daily operations. I naturally gravitate to high-scale backend challenges where latency budgets, failure behavior, and cloud cost are first-class design inputs. Outside build cycles, I show up at AI events and hackathons to pressure test ideas, sharpen execution, and meet people serious about building durable AI products.
2019 - 2022
FoundationStudent · Sri Chaitanya Techno School
High school at Sri Chaitanya Techno School gave me the science foundation I still lean on: PCM + Biology, pattern thinking, and a habit of asking why systems behave the way they do.
CGPA: 8.7
2022 - Present
FormationM.Sc. AI & ML Student · Coimbatore Institute of Technology
At Coimbatore Institute of Technology, I am pursuing an M.Sc. in Artificial Intelligence & Machine Learning while turning coursework into practical builds across ML, data systems, and full-stack AI.
CGPA: 8.67
2023 - 2024
LeadershipInnovation Lead · GDSC CIT & YI YUVA CIT
Google Developer Student Clubs, CIT
10+ AI/ML and web workshopsYI YUVA, CIT
Innovation challenges and mentorship programs2024 - Present
CommunityFounder / HR Manager · SDC CIT & The Snippet App
Student Developers Cell - CIT
30+ hackathons, speaker sessions, and code sprintsThe Snippet App
15+ developers and designers onboardedJun 2025 - Nov 2025
Industry AIML Engineer Intern · OptiSol Business Solutions
As an ML Engineer Intern at OptiSol Business Solutions, I built FastAPI services, AWS data and ML pipelines, SageMaker model workflows, Glue-powered feature engineering, and a React operations control panel for insurance AI workflows.
Skills / Toolchain
Organized by the kind of problem I am solving, from model work to backend systems and deployment plumbing.
Blogs / Idea Lab
These essays break down practical machine learning problems in plain language, while still respecting the engineering nuance behind production-grade AI systems.
Exploring boredom modeling, behavior signals, and personalization drift across streaming systems using ML and recommender strategy.
Read ArticleA practical take on why large language models often fail on figurative and culturally dense expressions, and what culture-aware context handling could look like in real NLP systems.
Read ArticleA deep dive into cognitive offloading, AI-assisted reasoning, and how humans are evolving into biological interfaces for machine intelligence in the prompt-driven era.
Read ArticleUpcoming deep dives: model interpretability, GenAI architecture trade-offs, and real-world ML system design patterns.
Follow on MediumProjects / Ship Log
From research-heavy models to production-grade platforms, each build is designed to solve real problems with clear engineering trade-offs and strong execution discipline.
A browser-only encrypted prompt and file sharing app with short links, password-protected capsules, burn-after-read delivery, and ciphertext-only storage.
Tech Stack: HTML, CSS, JavaScript, Web Crypto API, AES-256-GCM, PBKDF2, Upstash Redis, Vercel Blob, Vercel
A transcript-grounded product and growth research assistant with persistent chat, hybrid retrieval, validated citations, and markdown/HTML artifact generation.
Tech Stack: Python, FastAPI, PostgreSQL, pgvector, Ollama, Anthropic/OpenRouter, Docker
A production-grade security SDK for AI agents and MCP systems that checks prompts, tool calls, memory writes, documents, and outputs before risky content reaches production environments.
Tech Stack: Python, MCP, FastAPI/Flask integrations, JSONL auditing, optional ML detectors, security policy enforcement
An offline desktop editor for PDFs and Markdown documents that lets users open, inspect, edit, merge, redact, sign, and export files locally without uploading them to the web.
Tech Stack: Tauri 2, Vue 3, Rust, PDFium, Markdown editing, Windows integration
Developed an agent runtime monitoring platform with real-time trace streaming, span waterfalls, governance policies, cost forecasting, alert routing, PII-safe audit exports, and team budget controls.
Tech Stack: React 18, TypeScript, FastAPI, Python, SQLAlchemy, PostgreSQL, TanStack Query, Recharts, SSE, Docker
Designed a dual-module deep learning pipeline for motif prediction in genomic sequences with a Variational Autoencoder for synthetic sequence generation and a CNN classifier for binding property prediction.
Tech Stack: Python, TensorFlow, Keras, NumPy, Deep Learning
Built a multilingual sentiment dashboard with DistilBERT and KeyBERT, including threaded real-time processing and interactive Streamlit visualizations for regional sentiment tracking.
Tech Stack: Python, Streamlit, Transformers, DistilBERT, KeyBERT, Matplotlib
Developed an NL-to-SQL chatbot using Groq LLaMA3 and LangChain agents that handles nested logic, aggregates, and filters to make relational data accessible for non-technical users.
Tech Stack: Python, LangChain, MySQL, Streamlit, Groq API, Prompt Engineering
Created an ALPR system using YOLOv8 detection and Tesseract OCR, trained on a custom-annotated dataset and deployable as a lightweight Streamlit interface.
Tech Stack: YOLOv8, OpenCV, PyTorch, Tesseract OCR, Python, Streamlit
Designed a cross-platform planner using location, preferences, and dietary goals with offline SQLite tracking, RSS food updates, and personalized grocery recommendations.
Tech Stack: Flutter, Dart, SQLite, REST APIs, Riverpod, Email Auth, RSS Feeds
Architected a self-supervised transformer encoder with RevIN normalization and Conv1d patchification, scaling to process 10M+ time-series points and reducing reconstruction error through masked patch optimization.
Tech Stack: Python, PyTorch, sktime, Transformers, RevIN
Built a FastAPI-backed analytics assistant that turns natural language requests into Plotly charts with support for CSV, JSON, XLSX uploads and export-ready visuals.
Tech Stack: FastAPI, Python, Plotly, LLM Integration, JavaScript, HTML/CSS
Built a full-stack financial reconciliation platform with a hybrid matching engine combining rule-based logic, pgvector similarity search, and LLM tie-breaking for traceable and faster reconciliation.
Tech Stack: React, Vite, FastAPI, PostgreSQL, pgvector, SQLAlchemy, OpenAI API, Docker
Built a document sanitization pipeline for PII-heavy PDFs with OCR, entity extraction, and forensic redaction. The browser extension MVP intercepts PDF uploads, routes them through the backend sanitize flow, swaps in the cleaned file, respects an explicit allowlist, and surfaces a popup health and error dashboard.
Tech Stack: FastAPI, Python, PyMuPDF, pytesseract, Presidio, spaCy (en_core_web_trf), OpenRouter, rapidfuzz, Tailwind CSS, Vanilla JavaScript, PDF Redaction
Contact / Build Together
Whether it is a product idea, internship, collaboration, speaking invite, or a difficult engineering challenge, I am always open to meaningful conversations.
AI / ML Intake
Share the intent, constraints, and a link if you have one. Structured details help me reply with a useful next step.
GitHub
Portfolio
Location
Coimbatore, Tamil Nadu, India
Whether it is a quick question or a full build sprint, I am always excited to collaborate with people who care deeply about creating real impact.