Siddarth Bhave

- Led benchmarking and engineering efforts to integrate Mimir into firm's observability stack resulting in seamless integration with Grafana, increasing metric retention by 50% (+15 days), and decreasing query time for very-long-range queries from 5 minutes to 45 seconds

- Solely designed and developed a distributed, highly scalable ETL pipeline that extracted 75 million Kafka records per minute, transformed them into OpenTSDB format, and loaded them onto a highly available metrics aggregator, Cortex
- This increased real-time metric retrieval up to 60 days (+200%), reducing query time by 50%, and encouraging stakeholders to move towards modern observability, resulting in an early promotion and a tech-excellence award

- Revamped Windows server monitoring across the firm by seamlessly automating the integration of Prometheus exporter, which made Windows server monitoring highly available, reduced the system usage for monitoring by 50% and tripled the metric export rate for acute monitoring
- Collaborated with cross-functional teams to engineer automated target discovery and alerting which reduced the server outage alert time by 75% ( < 30s) and cut down turn-around-time by 20%

- Led a team of 15 junior developers and delivered 6 projects to our diverse customers in just 26 months, while being completely self-funded and profitable since day 0
- Architected the digitization of an ERP system for a 1,200-employee factory by automating tool management and tracking employee efficiency through cross-functional collaboration with industrial managers and employees, resulting in a 15% increase in employee efficiency and reducing the bookkeeping time by 75%
- Engineered a LangChain-based chatbot, modeled after Retrieval-Augmented Generation, enabling autonomous extraction of data from user corpus and processing natural-language user queries, resulting in improved customer response time by 20%

- Rearchitected pose classification using BlazePose and several CNNs on redrawn pose skeletons, increasing efficiency by 7% over state-of-the-art models.

- Developed a sentiment analysis model using GloVe vector embeddings and Graph CNN, boosting accuracy by 12% over existing analyzers.