AI Researcher · Princeton IT Services
Neelanjana Pal.
Ph.D., ECE
AI Researcher with a Ph.D. in Electrical & Computer Engineering and 6+ years of combined research and industry experience. Specialized in agentic AI systems, multimodal reasoning, and RAG, with deep expertise in uncertainty quantification, graph neural networks, and scalable AI architectures. Proven track record translating advanced theory into research-grade prototypes and enterprise-ready systems.
About
Bridging AI research and real-world systems
I work at the intersection of AI research and enterprise engineering: taking methods from formal verification, uncertainty quantification, and graph learning, and turning them into agent-based systems that hold up in the real world.
With 6+ years across research labs and industry — from proving neural networks safe during my Ph.D. at Vanderbilt, to production ML at MathWorks, to enterprise agentic AI today — I care about one question: when an AI system makes a decision, how much can you trust it?
Education
Ph.D. & M.S., ECE — Vanderbilt University
B.E.E. — Jadavpur University, Kolkata
Research Areas
Agentic AI · Multimodal Reasoning · RAG
Uncertainty Quantification · Formal Verification
Published At
CAV · FMICS · FMAS · ICMLA
Formal Aspects of Computing
Agentic AI Systems
Multi-agent architectures for enterprise workflows — autonomous reasoning, planning, and coordination with rigorous evaluation of performance, robustness, and scalability.
Retrieval & Reasoning
RAG pipelines, SLM architectures, graph-augmented retrieval, and hybrid multimodal search combining semantic embeddings with causal behavior signals.
Trustworthy AI
Uncertainty quantification, formal verification, adversarial robustness, and explainability — from provable guarantees in research to observability and governance in production.
Experience
Where I've worked
AI ResearcherCurrent
Jan 2026 – Present
Princeton IT Services, Inc.
- Designing a multi-agent architecture for mid-size enterprise workflows — document routing, cross-department search, and approval-chain automation — integrating MS Copilot and Claude-based agents into a unified orchestration layer.
- Leading a team of 2 on Claude-based agentic systems for enterprise clients; conducting applied research on autonomous reasoning, planning, and multi-agent coordination with evaluation metrics covering agent performance, robustness, and scalability.
- Architecting RAG pipelines and SLM architectures for scalable, real-world enterprise applications, validated through iterative design reviews with senior technical leadership.
- Designing data posture and telemetry infrastructure for agent observability — pipeline monitoring, behavioral tracing, and governance across multi-agent workflows.
AI/ML Researcher (Collaborative)Current
Oct 2025 – Present
Independent AI/ML Research & Prototyping
- Designing a hybrid multimodal retrieval architecture combining embedding-based semantic vector search with causal user-behavior signals for vibe-based product discovery.
- Improving cluster coherence by building LLM-powered product clustering pipelines integrating Elasticsearch full-text retrieval with graph-structured product relationships.
- Reducing cold-start errors in personalized recommendation by leveraging causal interaction histories as ranking features within a hybrid tag + embedding retrieval system.
Senior ML Engineer / Senior Software Engineer
Jan 2024 – Sep 2025
MathWorks
- Improved Copilot subsystem retrieval precision while reducing LLM token usage with a graph neural network-inspired heat-kernel diffusion approach (KG2RAG) on Simulink dependency graphs.
- Improved SLA adherence across four engineering teams by framing support-case distribution as a causal time-series problem — SHAP causal attribution on XGBoost/LightGBM models within an optimization framework.
- Designed adversarial training pipelines (FGSM, BIM, PGD) with uncertainty quantification checks across CV + time-series domains; developed Fourier Neural Operator architectures for forecasting under distribution shift.
- Deployed an internal OpenAI Chat Completions mock API cutting token spend; built a consumer product safety classifier with causal interpretability, and CV pipelines for ecological monitoring (Faster R-CNN, YOLO, U-Net, DeepLab).
Graduate Research Assistant
Aug 2019 – Dec 2023
veriVITAL Lab, Vanderbilt University
- Advanced uncertainty quantification for safety-critical AI by developing star-set reachability analysis for DNNs — provable robustness bounds at 20–1400× faster runtimes than state-of-the-art tools; published at CAV, FMICS, FMAS.
- Extended formal verification to multimodal neural architectures (semantic segmentation, LSTMs, U-Net, autoencoders) using relaxed reachability and zonotope pre-filtering.
- Applied SHAP-based causal feature attribution in DeepECO to identify energy-signature drivers of occupancy from smart-meter time-series.
- Led VNNComp 2020–2023 benchmarking; collaborated with Collins Aerospace and MathWorks on scalable verification tooling on AWS HPC with modular Docker codebases.
Intern, Deep Learning Verification & Validation
May – Sep 2022
MathWorks
- Shipped formal verification features in MATLAB R2022b, enabling engineers to apply uncertainty quantification and robustness checks to production AI models; collaborated with ETH Zurich, Vanderbilt University, and Collins Aerospace.
Project Engineer
May 2015 – Jun 2017
Rockwell Automation (Shanghai, China & India)
- Delivered automation solutions from design to commissioning for global clients (Hindustan Unilever, Colgate Palmolive, MARS), ensuring reliable large-scale system integration across manufacturing plants.
- Selected as 1 of 5 engineers nationally (India) and 1 of 14 globally — the only woman in the cohort — for a 3-month international training in Shanghai, presenting capstone outcomes to global leadership.
Teaching, Mentorship & Leadership
2024 – 2025
Mentor & Leadership · MathWorks
Created and delivered technical training on ML, Deep Learning, and Image Processing across US and UK offices; mentored MathWorks interns and Georgia Tech undergraduates.
2020 – 2021
Mentorship · Vanderbilt University
Supervised interns and undergraduate researchers in ML verification and GANs, expanding the number, variety, and quality of NNV evaluation case studies.
2017 – 2019
Teaching Assistant · Vanderbilt University
Mentored 60+ students across Programming, Digital Logic, Circuits, and AI Verification courses; Head TA for two years.
Publications
Selected publications
Robustness verification of DNNs using star-based reachability analysis with variable-length time series input
FMICS 2023
Formal verification of LSTM-based audio classifiers: A star-based approach
FMAS 2023
Robustness verification of semantic segmentation neural networks using relaxed reachability
CAV 2021
Verification of piecewise deep neural networks: a star set approach with zonotope prefilter
Formal Aspects of Computing 2021
DeepECO: Applying deep learning for occupancy detection from energy consumption data
ICMLA 2019
Projects & Open Source
Code that ships
Skills
Technical toolkit
Agentic Frameworks
ML / AI
Libraries & Tools
Languages & Platforms
Education & Service
Background
2023
Ph.D., Electrical & Computer Engineering
Vanderbilt University
2020
M.S., Electrical & Computer Engineering
Vanderbilt University
2015
B.E.E., Electrical Engineering
Jadavpur University, Kolkata
Awards
- Research Assistance Fellowship (ISIS, Vanderbilt, 2019–2023)
- Graduate School Travel Award (2023)
- Summit on Foundations of Data Science Travel Grant, San Francisco (2019)
- Summer School on Formal Techniques Student Grant (2021, 2022)
Reviewer
- Women in Machine Learning 2024
- Science of Computer Programming 2024
- CVPR, ICCV, AAAI, CAV, EMSOFT, ICCPS, HSCC (2020–2023)
Community
- Global Shapers Alumni Nashville (Curator & Vice-Curator, 2020–2023)
- Climate Reality Leader, Al Gore’s Climate Reality Project (2022)
Events & Speaking
Where you might have seen me
Grace Hopper Celebration (GHC)
Supported hands-on technical sessions at the world’s largest gathering of women and non-binary technologists.
International Women in Engineering Day
Presented on engineering careers and pathways for women in AI and engineering.
MATLAB Expo
MathWorks’ flagship conference on engineering, AI, and model-based design.
AWS Conferences & Summits
Cloud and AI infrastructure events — keeping current on scalable ML systems and agentic AI tooling.
CAV · FMICS · FMAS · ICMLA
Presented peer-reviewed research on neural network verification and applied deep learning at academic venues.
Summer School on Formal Techniques
Selected for the SRI International summer school on formal methods, both years on a student grant.
Summit on Foundations of Data Science
Attended the summit in San Francisco on a competitive travel grant.
Beyond the Work
The non-technical side
Research is what I do, not all of who I am. Away from the keyboard I write, wander, and work on things that outlast any one project.
Poetry & Writing
Words came before code. I write poetry and short reflections — on distance, belonging, and the small physics of everyday life. Some of it will find its way to the blog.
Read the personal blog →Photography & Travel
Mostly landscapes, mostly national parks. From sequoia groves to canyon rims — the photos on this site are my own, taken somewhere far from a lab.
Community & Climate
Curator & Vice-Curator of Global Shapers Nashville (2020–2023) and a trained Climate Reality Leader with Al Gore’s Climate Reality Project (2022).
Blog
Latest writing
Contact
Let's talk.
Open to research collaborations, speaking, and conversations about agentic AI, retrieval systems, and trustworthy AI. The fastest way to reach me is email or LinkedIn.