Sabita Kumari
AI Engineer / Research Assistant at Northeastern University
About
I'm Sabita Kumari, an AI Engineer and Research Assistant at Northeastern University. My journey in tech spans nearly two decades, starting in software engineering and evolving into a deep focus on Generative AI and LLMs. I specialize in building agentic workflows, RAG pipelines, and exploring the Model Context Protocol (MCP) to bridge the gap between AI and traditional backend systems. I’m passionate about 'learning in public' and believe that true AI engineering is about solving real-world problems with a solid foundation in system design. I love sharing my lessons from building end-to-end AI applications and am always looking to connect with fellow engineers and researchers to collaborate on production-grade AI and modern orchestration. Let's talk about trade-offs, automation, and building tools that actually reduce friction.
Networking
What I can offer
- ›Open-source code and technical blogs
- ›Lessons learned from building 7–8 AI applications
- ›Expertise in RAG pipelines and LLM fine-tuning
Looking for
- ›expanding my professional network
- ›exploring mutual opportunities in AI/ML and agentic systems
- ›collaboration on production-grade AI systems
Best fit for
Current Interests
Background
Career
Began as a Software Engineer in 2005, progressing through senior roles at Bentley Systems and Capgemini before pursuing an MS in Computer Science and transitioning into AI Engineering and Research.
Education
Master of Science in Computer Science, Northeastern University (2023 – 2025); Bachelor of Technology in Computer Science & Engineering, Biju Patnaik University of Technology, Odisha.
Achievements
- ›Achieved 100% syntax accuracy in Text-to-SQL fine-tuning using Mistral and LLaMA
- ›Reduced processing overhead for CT/MRI scans by 20% at Capgemini
- ›Built a custom AWS Deploy Agent using MCP and Claude for automated ECS deployment
- ›Optimized MySQL query execution time by 40% in software engineering roles
Opinions
- Being an AI Engineer is about solving one's own problems first rather than just collecting certificates.
- AI skills are insufficient without a strong grasp of traditional system design and backend fundamentals.
- System design is a series of trade-offs rather than a search for a single 'right' answer.
- Automation is addictive, but humans must stay involved in the 'struggle' of thinking.