Research & Projects
Published research in AI/ML and bio-energy systems, plus production engineering work in distributed systems and cloud architecture.
Publications
Peer-reviewed research contributions (2 papers)
AI-Driven Optimization of Bio-Energy Systems: Models for Resource Assessment and Emission Reduction
Authors: Rushali Rajaram Katkar, Sajal Suhane, Smita Suhane, S. Sugumaran, Santosh Bhauso Takale, Surekha Dehu Khetree, Shyamsing Thakur, Shital Yashwant Waware, Anant Sidhappa Kurhade
Systematic study on AI-driven models for optimizing bio-energy systems, focusing on resource assessment and emission reduction strategies using machine learning techniques.
Robolution: Real Time Predictive Analytics for Industrial Robots
Authors: Sajal Suhane, Pramod D. Patil, Ravi Mishra, Simran Koul et al.
Research on distributed predictive maintenance systems for industrial automation. Novel approach to distributed machine learning for real-time predictive analytics on IoT sensor networks.
Major Projects
Production systems and engineering contributions
Cloud-Native Platform Overhaul
Goldman Sachs – Associate
Led the transformation of a legacy Sybase IQ–based operational store into a cloud-native architecture, serving as Tech Lead for a 10-member team directing system design, roadmap planning, and cross-organizational stakeholder alignment.
Challenge
Legacy Sybase IQ operational store with on-prem storage dependencies, high costs, and slow query execution times.
Solution
- Architected scalable data platform using Kubernetes, Snowflake, Contour HTTPProxy, and event-driven ingestion pipelines
- Designed new Spring Boot microservice (migrated from Dropwizard) retaining all API contracts
- Seamlessly switched clients to Kubernetes-backed endpoints
Impact
- $1M+ annual cost savings
- 135% faster ingestions
- 650% faster query execution
- Eliminated on-prem storage dependencies
AWS Glue Migration Framework
Goldman Sachs – Analyst
Developed a reusable migration framework enabling teams to move on-prem Spark workflows to AWS Glue with minimal changes. Also mentored an intern and co-built an AI-powered knowledge chatbot.
Challenge
On-prem Spark workflows processing in 20+ minutes with high infrastructure costs. Support tickets taking 5 hours to resolve due to scattered documentation.
Solution
- Built reusable Spark → AWS Glue migration framework with minimal code changes
- Co-built AI-powered knowledge chatbot indexing nested Confluence spaces
- Built S3-backed intermediate computation system for Market Risk
Impact
- $230K annual savings per workflow
- 90% reduction in processing time (20min → 2min)
- Support ticket resolution: 5 hours → 1 hour
- 25% workflow efficiency improvement for Market Risk
Distributed Operational Store
Goldman Sachs – Intern
Engineered core components of a distributed operational store processing 3B+ events per second across Risk & Finance teams.
Challenge
Operational store handling billions of data points per second lacked observability and had ingestion timeout issues.
Solution
- Engineered core distributed store components for Risk & Finance
- Designed Kibana dashboards using live API metrics
- Built queuing-based ingestion POC eliminating timeouts
Impact
- 31.7% efficiency improvement
- 30% reduction in ingestion time
- Enhanced observability across Risk & Finance
Biomedical Sensor Data Platform
University of Texas at Dallas
Developed automated algorithms for cleaning and analyzing high-frequency biomedical sensor data, designed cross-platform applications for real-time sensor analytics, and optimized databases for wearable device data.
Challenge
Processing large volumes of biomedical sensor data requiring 0.6 FTE of manual effort. Millions of readings per minute from wearable devices needed efficient storage and real-time analytics.
Solution
- Developed automated algorithm for cleaning high-frequency biomedical sensor data
- Designed cross-platform applications for real-time sensor analytics
- Optimized databases to manage millions of readings per minute
Impact
- Reduced manual effort by 0.6 FTE
- 70% operational efficiency improvement
- Real-time processing of wearable device data