I take prototypes into production and keep them running there.
Thirteen years building software, three of them as CTO. These days that means retrieval that holds its accuracy on real archives rather than sample folders, agents that terminate and recover instead of looping, and voice pipelines that stay under a second end to end.
Based in Hyderabad. Working remotely with teams outside India since 2013.
Retrieval systems — chunking and parsing for messy real-world documents, hybrid and graph retrieval, reranking, grounding checks, and evaluation harnesses that tell you whether a change helped or hurt.
Agentic systems — bounded loops, tool contracts, retries and fallbacks, human handoff where it belongs, and traces detailed enough to reconstruct any run after the fact.
The platform underneath I build and own: web applications, cloud and system architecture, and the work that keeps a live product stable and hard to attack. Nine years running production SaaS, including keeping one standing through DDoS, spam and brute-force attempts.
Getting stalled builds moving — reading the code and the traces on a system that isn't working, and saying plainly what's wrong and what it costs to fix.
Architected an LLM-powered live translation system running under a second end-to-end, and shipped telephony agents handling real customer calls with context carried across the conversation.
Fraud detection system for bank transactions.
OCR - document parser to process invoices and process transaction automatically.
Chatbot for finance system to answer account related questions.
Quant finance - Multi-Agent system to analyse stock and predict Buy / Hold / Sell.
Before that, nine years with one group in Kuwait — the last three as CTO — owning engineering for two SaaS products. Moved the platform to microservices, and took web application performance up 10x through bundle and loading strategy, edge caching, and search offloading.
AI LangChain · LangGraph · ADK · RAG · Graph RAG · MCP · A2A
Haystack · Qdrant · vLLM · LiteLLM · AWS Bedrock · Claude · OpenAI
ML PyTorch · Tensorflow · Scikit-learn · BigQuery · AutoML
AWS SageMaker
Languages Python · Go · TypeScript · Rust
Data PostgreSQL · MongoDB · Redis · Qdrant
Infra AWS · Azure · GCP · Docker · Kubernetes · Terraform · Mlflow
Kubeflow
- Google Cloud — Generative AI Leader
- AWS Certified AI Practitioner
- Microsoft Certified: Azure AI Fundamentals
- Hugging Face — Fine-tuning Language Models · LLM Fundamentals · Fundamentals of MCP
Two-week diagnostics on stalled AI systems, project builds, and fractional AI lead work. Five to six hours of daily overlap with European hours.