About
I design and deploy real-world AI systems, LLM agents, retrieval pipelines, and ML platforms that scale in production and move business metrics. Research-driven, engineering-focused, impact-obsessed.
Blog Posts
Latest articles and insights from my Substack
Loading...
Loading blog posts...
Honors & Awards
Winner at Smart India Hackathon 2022
National level hackathon winner
Professional Experience
AI Developer @ Commvault
April 2025 - Present
- Support chatbot handling mission-critical queries with 60% deflection rate without escalations
- Built custom conversational AI framework from scratch, improving completion rate by 10%
- Enhanced RAG retrieval pipeline's MRR metric by 10% through 8+ customized experiments
- Integrating AI directly into Commvault's product and security assessment flows
- Engineered internal LLM system with fine-tuned open-source models (Llama, GPT-OSS), achieving 30% cost reduction
Software Engineer @ Druva
July 2023 - April 2025
- Developed a custom anomaly detection engine for Druva's cloud cost usage and metrics data, enabling proactive scanning and alerting
- This system tracks approximately 600,000 metrics, resulting in monthly savings of $10,000 in cloud costs and significantly reducing downtime through early mitigation
- Solely orchestrated the automation of Jenkins server deployment as an ECS service, implementing plugins for a seamless and efficient service deployment process
- Improving scalability to beyond 100% of current setup and saving $10K
- Engineering scalable Terraform managed infra, optimizing infrastructure to handle millions of user requests and 300+ services
- Enhanced Terraform and Jinja templates for improved AWS deployment efficiency
Enterpret
October 2022 - January 2023
Intern
- Implemented an automated and optimized selection process for the best model for production, taking into account client and global performance
- Cut production time by a factor of 10 with the introduction of a custom cost function for specific pipelines, resulting in tailored performance improvements for each customer
- Streamlined data generation and filtering pipelines to achieve a 13% increase in clean data
- Boosted the performance of the Cross encoder pipeline by 6x, leading to a significant overall product improvement
- Conducted 5+ experiments to enhance semantic similarity, training speed, and data imbalance handling, leading to 17% improvement
Prodigal
May 2022 - August 2022
Intern
- Redaction engine for customer service processing 100k+ calls daily
- Optimized production pipeline with 30% metric improvement
Solytics Partners
September 2021 - January 2022
Intern
- Built time series pipelines for financial institutions
- Delivered end-to-end training and deployment for macro/micro economic analysis
Solytics Partners
November 2020 - May 2021
Intern
- Led financial risk modeling platform module
- Deployed production scripts for Fortune 500 clients
- Optimized NLP pipeline inference 7x (used by HSBC)
Services
NLP
End to end nlp pipeline with deployment support for any cloud provider majorly AWS. And also serverless consultancy for inference.
Computer vision
Implementation of STOA papers for any of research pourposes or edge deployment using TF-lite
ML support
90% of production pipelines are ML based and sometimes a good SQL querry can also do the work.