Co-Founder & CTO | PhobosQ & k22.ai
I architect the infrastructure for autonomous enterprise intelligence. Operating at the convergence of advanced Machine Learning research and high-availability software engineering, I transition AI from stochastic, chat-based novelties into deterministic, stateful Agentic Systems capable of executing complex business logic at scale.
In an industry saturated with fragile prototypes, I specialize in building production-grade systems. My engineering philosophy prioritizes rigorous testing, multi-tenancy, and fault tolerance over mere capability.
"The future of enterprise is not just about data analysis; it is about autonomous execution."
The world's first and most advanced autonomous AI Data Scientist agent. Multi-tenant, secure, enterprise-grade. Autonomously navigates enterprise databases containing over 3,000 tables (SAP S/4HANA, Salesforce Health Cloud, Oracle) without hallucination.
View Product Features AI Data Scientist, Autonomous Agent, Enterprise AI, Multi-tenant, SAP, Salesforce, Oracle, PhobosQThe first foundational AI platform designed for 90 million Tamil speakers. Bridging advanced RL-based thinking (CoT, Parallel Tree) to truly democratize access to intelligence for farmers.
Tamil LLM, Foundational Model, High-reasoning AI, Language Model, Regional AI, RL, Chain of ThoughtAn autonomous agent leveraging 'Computer Use' capabilities to navigate web interfaces, analyze competitive metrics, and execute SEO strategies without human intervention.
AutoSEO, Computer Use, Autonomous Agent, SEO Automation, Web Automation, Enterprise SEOEnterprise document processing for Indian languages. Extracts entities, tables, and insights from complex documents with high accuracy.
Document Intelligence, OCR, Indian Languages, NLP, PhobosQTransforms Tamil and Indic language calls into actionable insights with 95% accuracy. Features real-time sentiment analysis, sales opportunity detection, and coaching to boost conversion by 35%.
Call Analytics, Indo-Aryan Languages, Voice AI, Sentiment Analysis, Sales OptimizationAn enterprise-grade Retrieval-Augmented Generation system with multi-LLM support, hybrid vector search, computer vision, and optimized document processing for scalability and performance.
RAG, LLMs, Vector Search, Qdrant, YOLOv8, Python, StreamlitAn enterprise-grade AutoML studio. Users can upload datasets, choose target variables, and build models. It's scalable, supports on-premise deployment, and uses Transformers.js with WebGPU for acceleration.
AutoML, TensorFlow.js, Next.js, Transformers.js, WebGPUAn enterprise-grade, multi-agent Text-to-SQL system for complex databases using a hybrid MCTS approach. Built with 5 LLM agents for high accuracy.
Vite, React, Shadcn UI, Multi-agent Systems, Text2SQL, LLMA Streamlit Cloud application that takes a startup idea and provides expert-level guidance on next steps, funding strategies, and more.
Streamlit, Python, Generative AI, Startup AdvisoryAn OpenAI-like playground for Groq models. Allows users to interact with Groq models, use their API keys, and test custom prompts. Supports on-premise installation.
Next.js, Radix UI, Shadcn UI, Groq API, LLM PlaygroundProduction-grade RAG system with NVIDIA NIM and LlamaIndex for enterprise-scale document processing. Features multi-LLM support, parallel processing, and robust security.
RAG, LlamaIndex, NVIDIA NIM, Python, Streamlit, Distributed SystemsA collection of my early machine learning projects, including: 1. A Disease Prediction System 2. Credit Card Fraud Detection, and 3. Real-time Face Detection.
Python, Scikit-learn, TensorFlow, OpenCVDeveloped a vehicle recognition system for speed, OCR, and in/out counts. Deployed on Google Coral Edge-TPU and Jetson, increasing processing efficiency by 40%.
Computer Vision, YOLOv8, TensorRT, Edge-TPU, OpenCVDeveloped an interactive system for segregating newborn chicks in farms using Quantum-SVMs, achieving a 30% reduction in classification time. Deployed on NVIDIA Jetson Nano.
Quantum SVM, CNN, PyTorch, Jetson Nano, MXNetImplemented a generative adversarial framework for bacterial behavior simulation, using GANs for data generation and RL for decision-making. Achieved 20% accuracy improvement.
Reinforcement Learning, GANs, PyTorch, TensorFlow, NLP638 contributions in the past year, focusing on solving fragmentation in the AI stack.
Research implementation of Monte Carlo Tree Search (MCTS) for Text2SQL, applying game-theoretic search algorithms.
Enterprise RAG system integrating YOLOv8, allowing agents to 'see' and correlate visual data within complex documents.
Active contributor ensuring the next generation of AI interfaces is robust and developer-friendly.
As Co-Founder & CTO of PhobosQ, I'm architecting the next generation of enterprise intelligence: autonomous AI agents that don't just answer questions—they execute complex business logic across production databases.
Deep CRM integration (Salesforce, SAP) that actually understands context and takes action.
AI-native technical interviewing platform for 100k+ concurrent candidates using WebRTC, SSE, and distributed caching.
The first AI platform for 90M Tamil speakers with cultural & linguistic adaptation at the model level.
Autonomous agent leveraging 'Computer Use' for competitive analysis and automatic content generation.