Dinesh Bastin
Senior Python Developer · AI Engineer · Chennai
Building intelligent systems with Python, LLMs & RAG pipelines — 6 years turning complex data into production-grade AI solutions.
What I work with
What I've been building
Natural language to SQL engine for warehouse ops. RAG-based architecture answers inventory, orders, picking and analytics queries in plain English against live databases.
Real-time dashboards with automated insight generation, anomaly detection and predictive analytics using historical data and ML models.
AI-powered candidate evaluation using ranking similarity algorithms and sentiment analysis from video interview responses to assist recruiters in shortlisting.
Marketing lead gen pipeline using microvertical segmentation and workflow automation with n8n to support automated outreach campaigns.
Research into orchestration frameworks, probabilistic reasoning and autonomous workflow automation. Exploring how agents collaborate on complex tasks.
GPT-4 as both SQL generator and judge. Detects hallucinated SQL, tracks token cost, and visualizes confidence metrics — inspired by the UQLM library.
View on GitHub →Detects real vs AI-generated images using a fine-tuned Vision Transformer. FastAPI backend, React web dashboard, Flutter mobile app and forensic heatmaps.
View on GitHub →Open-source prompt engineering tool for LLMs. Supports few-shot, Chain-of-Thought / Tree-of-Thought reasoning, RAG context, OpenAI and local Ollama models.
View on GitHub →Thoughts & tutorials
Most posts say quantization reduces memory. True. But we tested all three inside a retrieval-heavy production-style RAG pipeline — not toy prompts, real constraint-based responses. 6-bit is the sweet spot. Here's why.
The person behind the code
I'm Dinesh Bastin, a Senior Python Developer based in Chennai with 6 years of experience building software — and the last few years increasingly focused on AI, LLMs, and intelligent automation.
My background in Electronics & Communication Engineering gave me a strong foundation in systems thinking. That carries into every project — I care about how things fit together, not just how they work in isolation.
I've built across warehouse intelligence, analytics dashboards, HR tech, and marketing automation — always with Python at the core, and increasingly with RAG, NLP, and multi-agent frameworks as the intelligence layer.
Let's build something
Open to interesting projects, collaborations, and conversations about AI systems. Drop a message and I'll get back to you.