KYRON
An autonomous AI system that takes a business goal and figures out how to execute it — end to end.
View Project
KYRON is a full-stack SaaS platform where you define a business goal and an AI agent network handles the execution. It uses a multi-agent architecture — one agent plans, one monitors progress, one handles recovery when things go wrong. Every decision made is logged and fully auditable.
LLM Engine
A transformer-based language model built completely from scratch — not fine-tuning, not an API. The real thing.
View Project
Built a transformer-based LLM in PyTorch from the ground up — including a custom tokenizer, embedding layers, multi-head self-attention, positional encoding, and a complete training pipeline.
Self‑Healing RAG
A RAG pipeline that detects when it's wrong and automatically fixes itself — no silent hallucinations.
View Project
Most RAG systems return confident wrong answers. This one doesn't. After generating a response, it validates the output against retrieved context. If confidence drops below threshold, it rewrites the query and tries again.
Jal Vaani
A data intelligence platform for tracking, analyzing, and predicting groundwater levels across India.
View Project
End-to-end app built to handle water data at scale. Ingests real-time WRIS + NASA readings, runs ML models for anomaly detection and prediction, and surfaces insights through a visual dashboard.
SACH-AI
Multimodal AI that detects deepfakes, fake messages, AI-generated content, and cloned voices — and explains exactly why.
View Project
SACH AI analyzes text, images, videos, audio, and screenshots to determine if content is authentic or manipulated. It returns a trust score, risk level, and a clear explanation of the reasoning — not just a label.