Building AI products, LLM systems, and workflow intelligence tools
Sydney, Australia
I’m an AI builder focused on turning messy real-world inputs into structured systems that people can actually use.
Right now, I’m building Orchestra through Arrayah and the SH1P 1 Australia Cohort in Sydney. It is an AI-powered software intelligence layer and Product Brain for software teams, designed to turn messy project requirements into structured scope, execution plans, and live delivery visibility.
I’m currently pursuing a Master of Data Science and Innovation at UTS, and I’m especially interested in:
- LLM systems
- RAG pipelines
- multi-agent workflows
- applied AI products
- workflow intelligence for software teams
A private product I’m currently building through Arrayah.
What it is
- An AI-powered software delivery intelligence layer for managers and engineers
- Turns messy requirements into structured scope, execution plans, and live delivery insight
- Bridges the gap between business context and technical execution
- Focused on clarity, alignment, and execution visibility inside software teams
Sydney, Australia | 2026 – Present
- Currently building Orchestra through Arrayah
- Working on AI systems for requirement translation, execution planning, and software delivery visibility
Sydney, Australia | Feb 2026
- Built an end-to-end AI-powered resume ranking system
- Designed a hybrid pipeline with section-aware scoring, dynamic weighting, domain similarity analysis, explainable score breakdowns, and privacy-first local data handling
Remote | Dec 2024 – Jan 2025
- Contributed to AI-driven voice-to-voice translation systems
- Worked across NLP, speech recognition, and machine learning to improve real-time translation accuracy and contextual understanding
Bengaluru, India | May 2024 – Oct 2024
- Built applied AI systems for efficiency and safety
- Worked on conversational AI for software testing FAQs, terrain-adaptive vehicle intelligence, object detection for safety compliance, and Python automation workflows
Enterprise RAG, agentic workflow, and governance platform
- Full-stack AI product with FastAPI, Next.js, PostgreSQL, pgvector, Redis, JWT/RBAC, Docker Compose, and CI
- Implements document ingestion, vector retrieval, citation-grounded answers, confidence scoring, human review, audit logs, analytics, and evaluation workflows
- Designed to demonstrate production-minded GenAI engineering: security boundaries, provider abstraction, mock/local demos, observability, and deployment planning
Intent-Aware Data Science Copilot
- Takes a user’s analysis goal in plain English and runs end-to-end analysis
- Profiles and cleans data, generates explanatory visualizations, trains suitable ML models, and returns clear answers with supporting insights
- Built with FastAPI, ContextUI, scikit-learn, and LLM-based workflows
Local-first AI resume screening
- Ranks resumes using section-aware analysis instead of naive keyword matching
- Uses dynamic weighting across experience, projects, skills, and certifications
- Adds explainable scoring, clustering, and contextual insights while keeping raw data local
AI decision-support system for agriculture
- Built for the Mistral Hackathon in Sydney
- Explores district-level climate, crop viability, water usage, fertilizer impact, and policy-oriented agricultural insights
- Designed to support better agricultural decision-making with AI-powered recommendations
End-to-end AI ad generation workflow
- Generates marketing copy, CTAs, and AI-assisted video creatives across funnel stages
- Centralizes feedback, regeneration, and creative management in one workflow
- Built to speed up ad production and iteration
AI study companion grounded in class material
- Generates quizzes with explanations and citations
- Builds day-by-day study plans up to the exam date
- Designed for grounded revision from uploaded notes, slides, and PDFs
Voice-first AI travel companion
- Built for backpackers exploring Australia
- Combines voice input/output, itinerary planning, travel utilities, location context, and recommendations
- Focused on practical on-the-go travel assistance
Claude-powered candidate research agent
- Built for an executive-search style workflow
- Researches public candidate context and generates structured recruiter briefings
- Uses tool-assisted research, validation, and strict output formatting
Real-time vehicle intelligence
- CNN-based terrain detection for real-time vehicle behaviour adjustment
- Dynamically adapts suspension, braking, and ride height using sensor and camera data
- Developed during my time at Continental Automotive
- Dual Speaker Conversational AI Assistant
- PDF Chatbot / RAG Assistant
- AI-Powered Software Testing FAQ Chatbot
- Helmet Detection System
- Excel Automation Suite
- Safety compliance object detection workflows
- Python automation tools for repetitive operational tasks
- Submitted a technical paper on AI-Enhanced Terrain-Adaptive Vehicle Control System for the SAEINDIA International Mobility Conference 2024
- Showcased the concept and implementation at Continental Innovation Day
- Published “Real-Time Biometrics-Based Smart EVM with FPGA Implementation” in the International Journal of Scientific Research and Engineering Trends
- Built AgriSmart as part of the Mistral Hackathon in Sydney
Languages
Python, SQL
Core Areas
Generative AI, LLMs, NLP, RAG, Neural Networks, Data Visualization
Frameworks / Libraries
LangChain, FastAPI, Streamlit, TensorFlow, PyTorch, Hugging Face, scikit-learn, Pandas, NumPy
Tools
Git, GitHub, VS Code, Cursor, Claude Code, Notion, ContextUI, Codex
Master of Data Science and Innovation
2025 – 2027
Bachelor of Engineering in Electronics and Communication
2020 – 2024
- Email: karthikramesh2012@gmail.com
- LinkedIn: karthik-ramesh-2b52ab328
- GitHub: KarthikRamesh9149
I’m interested in AI products, startup ideas, workflow systems, and real problems that need strong technical execution.