From predictive statistical modeling to state-aware autonomous AI agent networks, my mission is to build software that doesn't just process data - it reasons, executes tools, and drives decisions independently.
Rohit's journey in Artificial Intelligence began over two decades ago. In 2004 at IIT Roorkee, under the mentorship of Dr. Sunil Padhi (HOD, Electrical Department), he completed his inaugural AI and Artificial Neural Network (ANN) project, writing custom neural network backpropagation algorithms ground-up in C Language to predict annual solar sunspot cycles.
In 2007 at Xalted Information Systems in Bengaluru, under the mentorship of Srinivas Sir, Rohit authored his first automation scripts in Unix shell scripting to parse, analyze, and process massive Call Detail Records (CDR). Ever since, he has cultivated a deep passion for playing with data, building high-throughput data engineering pipelines, and designing intelligent automation systems.
Over the past two decades, Rohit's expertise evolved alongside AI technology - moving from low-level C neural algorithms and Unix shell data automation to predictive deep learning, custom RAG pipelines, and state-aware Agentic AI networks.
As an independent AI consultant at rcode.in, Rohit helps organizations architect resilient multi-agent systems where AI handles complex enterprise workflows autonomously.
"From writing neural networks in C at IIT Roorkee in 2004 to orchestrating multi-agent networks today, the core principle remains: building AI that reasons and delivers real ROI."
Built custom backpropagation neural networks ground-up in C language under the mentorship of Dr. Sunil Padhi (HOD, Electrical Department) to predict annual solar sunspot cycles at IIT Roorkee.
Authored first automation scripts in Unix shell scripting under the mentorship of Srinivas Sir at Xalted Information Systems in Bengaluru for CDR processing, igniting a lifelong passion for data automation.
Architected automated ETL feature stores, linear/logistic classification models, and decision tree classifiers for enterprise datasets.
Scaled distributed ML pipelines processing millions of rows using Hadoop, PySpark, and SQL analytical data warehouse engines.
Transitioned to PyTorch deep learning, building autoencoder neural networks for high-frequency transaction fraud and anomaly detection.
Developed hybrid time-series demand forecasting models combining PyTorch LSTM neural networks and XGBoost gradient boosting.
Built specialized medical NLP search pipelines querying 4.5M PubMed articles using domain-tuned BioBERT embeddings and vector search.
Pioneered early RAG architectures connecting enterprise document vector stores with foundation models for context-grounded Q&A.
Combined Qdrant dense vectors, BM25 sparse keyword retrieval, Cohere reranking, and Neo4j entity graphs to eliminate RAG hallucinations.
Deployed self-hosted open-weight LLMs (Llama 3 70B) on private NVIDIA GPU clusters with vLLM PagedAttention and 62% API cost savings.
Architected state-aware multi-agent networks (LangGraph, CrewAI, AutoGen) with persistent graph states, tool sandboxes, and safety guardrails.
Leading rcode.in AI Consultancy, guiding CTOs on Agentic AI roadmaps, air-gapped local model serving, and enterprise ROI optimization.
No middle managers or generic account reps. Direct technical advisory from day one.
+91-7579-1857-75 | rohit@rcode.in
Book Direct Consultation