ML Software Engineer @ Rocket Software | ML Research Consultant @ UCSD Chiba Lab | Reinforcement Learning · Robotics · Agentic AI

Machine Learning Software Engineer at Rocket Software building AI-powered enterprise tooling (Rocket EVA™, SmartChat™), and Machine Learning Research Consultant at UC San Diego's Chiba Lab (RUBI Lab) publishing on reinforcement learning, multi-agent robotic behavior, and neurorobotics.

At Rocket, I own core features of the EVA™ AI platform (RESTful API, OIDC auth, MCP tool execution, Docker deployment), built a DSPy-powered SAS-to-Python code translation agent, and lead AI enablement training for 1800+ engineers globally. Won "Most Innovative" at the 2025 Rocket.Build Hackathon for an LLM-powered Scrum assistant.

At UCSD, I am 2nd author on an IEEE ICDL 2025 publication on explore-exploit dynamics in rat-robot interactions. I implement RL algorithms in PyTorch over ROS2 with sim-to-real transfer to physical robots, and I was founding engineer of a multi-million dollar AI-powered classroom embedded system.

Passionate about deep RL, sim-to-real transfer, multi-agent systems, and agentic AI. Published across IEEE, Frontiers in Psychology, and BCI. Open to mid-career opportunities and graduate programs in robotics and reinforcement learning.

1. Reinforcement Learning
2. PyTorch
3. Agentic AI / LLM Applications

Machine Learning Software Engineer III
September 2024 – Present | Waltham, MA

• Own core features of Rocket EVA™ AI platform: sanitized RESTful API, end-to-end OIDC authorization, streaming telemetry, async MCP tool execution engine, and structured Docker deployment for company-wide release.
• Built SAS-to-Python code translation tool using DSPy ReAct agent, collaborating on Agentic Context Engineering (ACE) optimization framework; created multi-agent workflows for architecture design, test generation, and automated code execution refinement with MLflow evaluation.
• Designed and implemented novel LLM Citation Generator for SmartChat™ using multi-Gaussian clustering algorithms and dynamic content similarity techniques, yielding 80% more relevant document references.
• Developed proof-of-concept few-shot LLM framework to translate natural language queries into executable SQL with regex-based syntax validation.
• Create and deliver AI foundations lecture series for company apprenticeship program at global offices, achieving 100% content application rate; leading coordination of enterprise-wide AI enablement program for 1800+ engineers.
• Won "Most Innovative" at 2025 Rocket.Build Hackathon (1 of 6 awards from 650+ global projects) for LLM-powered Agile assistant and MCP Server optimizing Jira workflows.

Machine Learning Research Consultant
September 2024 – Present | San Diego, CA

• Architected composable spatial and neural signal analysis pipeline and presented IEEE ICDL 2025 publication (2nd author) on explore-exploit dynamics in rat-robot social interactions.
• Developing bio-influenced deep RL regulation framework in MuJoCo simulation using multiple policy optimization algorithms to model exploration-regulation dynamics in autonomous robotic behavior.
• Founding engineer hired for multi-million dollar AI-powered classroom project; designed PTPv2-synchronized multi-modal embedded system with Arduino firmware for OpenBCI biophysical sensors and Basler camera integration.

Staff Machine Learning Research Engineer
January 2021 – September 2024 | San Diego, CA

• Spearheaded implementation of Reinforcement Learning algorithms in PyTorch for robotic PiRat over ROS2, utilizing data from 2500+ multi-agent experiments to model interaction dynamics.
• Developed live multi-agent pose estimation system using SLEAP neural networks with multi-threaded architecture.
• Built real-time Homebase Detection Algorithm with convolution feature extractor; autonomous behavioral routines simulated and transferred via sim-to-real pipeline to physical PiRat velocity commands.
• Enhanced and scaled production Computer Vision pose estimation software in C++ achieving 300% processing speed improvement.
• Developed signal processing applications for neural and physiological data (HRV, PPG) analysis.
• Co-authored Frontiers in Psychology 2022 publication and contributed to BCI 2025 abstract.

Data Science Teaching Assistant (COGS 108)
September 2022 – December 2023 | San Diego, CA

• Created weekly section presentations, podcasted to 800+ students, covering content review and interactively guiding students through programming assignments via live debugging.
• Formulated weekly exams, held tutoring sessions, scripted automated grading pipelines to optimize multi-LMS system.
• Instructed on systematically leveraging Exploratory Data Analysis techniques to answer complex data-focused questions.

President & Recruitment Advisor — Zeta Alpha Chapter
September 2021 – December 2023 | UC San Diego

• Led 30+ member chapter of National Honor Fraternity; increased membership by 250% over two recruitment cycles.
• Represented chapter at 2022 PSP National Convention in Washington, DC.
• Managed quarterly scholarship, fellowship, service, and fundraising programming.

• Explore-Exploit Behaviors During Rat-Robot Interactions Optimize Social and Spatial Security — IEEE ICDL 2025 (2nd author)
• Interactive neurorobotics: Behavioral and neural dynamics of agent interactions — Frontiers in Psychology 2022
• Robotic Exploratory Control Via Subcortical Oscillations — 11th International BCI Meeting 2025 (Abstract)

UC San Diego
Bachelor of Science, Machine Learning and Neural Computation | Minor: Computer Science | Concentration: Philosophy
September 2020 – December 2023
Major GPA: 3.91/4.0
Awards: Provost Honors (Warren College)

• Website: https://www.raunitkohli.com
• Google Scholar: https://scholar.google.com/citations?user=abZZ14QAAAAJ
• ICDL 2025 Paper: https://ieeexplore.ieee.org/document/11204421