Dipankar Chakraborty

Built and orchestrated multi-agent systems using frameworks like LangChain, CrewAI RAG, using vector databases, embeddings, and retrieval optimization. Enhance coding capabilities across various programming languages using high-quality training data for Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), LORA and QLORA. Develop data-driven solutions for function calling and agent workflows utilizing APIs across platforms such as Azure, iOS, GCP, AWS. Provide comprehensive post-training support to optimize models using techniques like SFT, Multilingual Fine-Tuning (MFMT), RLHF, and Direct Policy Optimization (DPO). Measure the performance of LLM using performance metrics such as BLEU, ROUGE, MAUVE, SacreBlEU, etc.

Built scalable ETL pipelines to ingest and transform data from CMS, student portals, and digital learning platforms. Developed analytics-ready datasets and standardized KPI frameworks to track student performance, engagement, retention, and feature adoption. Automated reporting workflows and Tableau dashboards, reducing average feedback incorporation time by 60%. Implemented data quality checks and pipeline monitoring, improving reporting accuracy by 35%. Optimized SQL queries and transformation jobs to improve dashboard refresh and reporting turnaround time by 20%.

Developed interactive data visualization tools and dashboards to communicate insights to stakeholders using JIRA. Collaborated with product development teams to identify data-driven improvements for content and user experience. Conducted A/B testing and statistical analysis to evaluate the effectiveness of platform features. Identified trends in student behavior resulting in 23% better student engagement. Analyzed teacher engagement and reduced content reorganization by 25%. Developed dashboards to track errors, reducing conceptual error reporting by 86%.

Segmentation of students and teachers based on plan, academic rigor and other criteria. Used Tableau to communicate the performance of teachers and students. Analyzed Formative assessment datasets leading to a 20% improvement in students' overall performance on Summative assessments. Supervised faculty performance and offered constructive criticism. Integrated structured and unstructured data from various internal and external systems.