Research Engineer · AI/ML · Time-Series Foundation Models
Vinamra Baghel
I build intelligent systems at the intersection of machine learning, time-series modeling, and agentic AI. Currently at IBM Research.
01 / Focus
What I work on
Time-Series Foundation Models
Foundation models for time-series data — how they are trained, evaluated, and made data-efficient, and what it takes for them to generalize across domains they were never shown.
Agentic AI
Systems that reason over streaming data: using tools, retrieving context, and maintaining memory so that decisions hold up as new observations arrive.
Intelligent ML Systems
Research and engineering across representation learning, synthetic data, and uncertainty modeling — and the systems work of turning those ideas into things that run.
02 / Selected
Selected work
Time-Series Semantic Intelligence Agent
An agent that reasons over live industrial time-series streams — detecting anomalies, searching semantically, and investigating root cause as data arrives.
Agentic AI · Streaming2025Data Evaluation Framework for Time-Series Foundation Models
Selecting the pretraining data that actually matters for forecasting foundation models — influential, diverse samples over sheer volume.
Time-Series · Evaluation2026Targeted Synthetic Data Generation
Closing specific gaps in a pretraining corpus with synthetic time series aimed at where the model is weakest — not bulk augmentation.
Synthetic Data · Time-Series2026Spatio-Temporal CO₂ Flux Estimation from Remote Sensing
Estimating high-resolution CO₂ emission flux from sparse satellite and ground-level proxies, using deep spatio-temporal models.
Climate · Remote Sensing2024
03 / Now
Now
Updated August 2026
- Building a Time-Series Semantic Intelligence Agent — an agent that reasons over live industrial data streams for anomaly detection, semantic search, and automated root-cause analysis.
- Working with streaming pipelines, time-series foundation models, and a knowledge-graph context layer, with the agent's tools exposed over MCP.
- On the research side: data-centric pretraining for time-series foundation models — selecting the data that matters and generating synthetic data to close the gaps.
04 / Path
Experience
- IBM Research
- Research EngineerTime-series foundation models, agentic AI over streaming data, and applied ML for climate and sustainability.2024 —
- IIT Bombay
- Dual Degree (B.Tech + M.Tech), Electrical EngineeringM.Tech specialization in Communication & Signal Processing; minor in AI & Data Science (C-MInDS).2019 – 2024
- Morgan Stanley
- Summer Technology AnalystInstitutional Securities Technology — a classification engine for filtering flaky tests in a large CI pipeline.2022
05 / Notes
Writing
Notes and essays are in progress. Planned topics include:
- Time-series foundation models
- Agentic AI
- Streaming ML systems
- Evaluating foundation models
- Building intelligent systems
06 / Contact