Janakiraman (Ram) Balachandran

I am a Ph.D.-trained mechanical engineer with 10+ years of experience building AI for real-world physical systems across water, renewable energy, oil & gas, and manufacturing. I am currently the Chief Data Scientist at SpiralData / Transient Detect in Adelaide, Australia, where I lead an AI platform that detects pressure transients (water hammer) in water distribution networks so utilities can prevent pipe failures and extend asset life.

Before that, I led machine-learning work on price and promotion optimisation at Complexica, automated property valuation at PointData, and AI sewer inspection (Dragonfly) at Hitachi Vantara. My research career spans Shell, where I built graph neural networks and physics-constrained ML for materials discovery, a postdoc at Oak Ridge National Laboratory, and a Ph.D. at the University of Michigan.

I am interested in physics-informed machine learning, time-series & signal processing, computer vision, and LLM systems (RAG & agents). I like using engineering intuition to build models that make assets safer, more reliable and more efficient, and turning unclear goals into scoped roadmaps that deliver measurable business impact.

Impact

Experience

• 2023 – Present · Adelaide, Australia
  Chief Data Scientist, SpiralData / Transient Detect
  • AI pressure-transient detection. Built an AWS-native platform that ingests >100 Hz pressure feeds and surfaces transient-risk views within ~15 minutes, giving water utilities visibility they never had before.
  • Designed detection and clustering algorithms with >97% reliability that tie invisible water-hammer spikes to asset failure, extending asset life by 5–10 years at hot spots.
  • Worked with Greater Western Water to find and mitigate transient sources, unlocking roughly $3M in avoided reactive work and underpinning SpiralData’s $1.5M scale-up capital raise.
  • Embedding AI into BAU workflows. Built a multi-modal ingestion pipeline (parsers, speech-to-text, embeddings, vector search, NL2SQL) so every answer cites governed data; proved it on a RAG use case surfacing financial-health signals for insurance risk modelling.
• 2022 – 2023 · Adelaide, Australia
  Lead Machine Learning Scientist, Complexica
  • Replaced spreadsheet-only trade planning for Endeavour Group with a guided AI workspace that captures objectives, constraints and deals before they reach retailers.
  • Fused POS, pricing, promo calendars and product hierarchies to model uplift, elasticity and cannibalisation at store–SKU–week depth, with optimisation layers recommending mechanics, depth, timing and price within business guardrails.
• 2021 – 2022 · Adelaide, Australia
  Data Scientist, PointData
  • Delivered parcel-level automated valuations with up to 90% accuracy for 200,000+ South Australian properties, shortening lender valuation cycles and helping councils find subdivision-ready lots.
  • Built pipelines ingesting sales records, cadastral parcels, zoning, GNAF addresses and auction feeds so the AVM always sees the freshest signals.
• 2019 – 2021 · Bangalore, India
  Data Science Manager, Hitachi Vantara
  • Founding member of Dragonfly, an AI sewer-inspection product built with Jacobs Water that turns raw CCTV into prioritised maintenance actions, trimming inspection and rehab cost by roughly 25%.
  • Trained deep-learning models on 300,000+ metres of footage, reaching >95% PACP coding accuracy in gritty sewer conditions.
• 2017 – 2019 · Bangalore, India
  Researcher, Shell
  • Led graph convolutional networks with multi-task learning to predict properties of 48,000 materials for energy conversion and storage (NeurIPS 2018 workshop).
  • Developed physics-constrained ML (bootstrapped projected gradient descent) giving simple, interpretable, transferable models from small datasets for supercapacitor design (Chem. Mater. 2019).
  • Built data-extraction and analytics workflows that identified new electrocatalysts for hydrogen-peroxide production, and led deep-learning classification of crystal structures from electron microscopy.
• 2015 – 2017 · Oak Ridge, TN, USA
  Postdoctoral Research Fellow, Oak Ridge National Laboratory
  • Built experimentally benchmarked ab initio models revealing how disorder, doping and strain govern ion transport in oxides, yielding design principles for fast-ion conductors for fuel cells and batteries.
  • Developed a materials-informatics framework combining massively scalable workflows, ab initio modelling and analytics, producing one of the largest defect datasets for perovskites (J. Phys. Chem. C 2017).
  • Applied data mining to extract synthesis parameters into a materials-synthesis database (J. Appl. Phys. 2018).
• 2009 – 2014 · Ann Arbor, MI, USA
  Doctoral Candidate, University of Michigan
  • Designed experimentally benchmarked models of nanoscale materials for thermoelectric devices (JACS 2011, APL 2012), explaining how material structure drives these properties (JPCL 2012) and enabling efficient high-throughput calculations (JPCL 2013).

Selected Publications

See the full list or my Google Scholar profile.

• Machine learning constrained with dimensional analysis and scaling laws: simple, transferable, and interpretable models of materials from small datasets
  N. Kumar, P. Rajagopalan, P. Pankajakshan, A. Bhattacharyya, S. Sanyal, J. Balachandran, U. V. Waghmare
  Chemistry of Materials, 31(2), 314–321 (2019)
  Featured in Chem. Mater. virtual issue on ML discoveries
  DOI
• Nanoscale control of oxygen defects and metal–insulator transition in epitaxial vanadium dioxides
  Y. Sharma, J. Balachandran, C. Sohn, J. T. Krogel, P. Ganesh, L. Collins, A. V. Ievlev, Q. Li, X. Gao, N. Balke, O. S. Ovchinnikova, S. V. Kalinin, O. Heinonen, H. N. Lee
  ACS Nano, 12(7), 7159–7166 (2018)
  DOI
• Data mining for better material synthesis: The case of pulsed laser deposition of complex oxides
  S. R. Young, A. Maksov, M. Ziatdinov, Y. Cao, M. Burch, J. Balachandran, L. Li, S. Somnath, R. M. Patton, S. V. Kalinin, R. K. Vasudevan
  Journal of Applied Physics, 123(11) (2018)
  DOI
• Defect genome of cubic perovskites for fuel cell applications
  J. Balachandran, L. Lin, J. S. Anchell, C. A. Bridges, P. Ganesh
  The Journal of Physical Chemistry C, 121(48), 26637–26647 (2017)
  DOI
• End-group-induced charge transfer in molecular junctions: effect on electronic-structure and thermopower
  J. Balachandran, P. Reddy, B. D. Dunietz, V. Gavini
  The Journal of Physical Chemistry Letters, 3(15), 1962–1967 (2012)
  DOI
• Effect of length and contact chemistry on the electronic structure and thermoelectric properties of molecular junctions
  A. Tan, J. Balachandran, S. Sadat, V. Gavini, B. D. Dunietz, S.-Y. Jang, P. Reddy
  Journal of the American Chemical Society, 133(23), 8838–8841 (2011)
  DOI
• Potential energy surface prediction of alumina polymorphs using graph neural network
  S. Sanyal, A. K. Sagotra, N. Kumar, S. Rathi, M. Krishna, N. Somayajula, D. Palanisamy, R. R. Ratnakar, S. Sanyal, P. Talukdar, U. Waghmare, J. Balachandran
  arXiv preprint (2023)
  arXiv
• MT-CGCNN: Integrating crystal graph convolutional neural network with multitask learning for material property prediction
  S. Sanyal, J. Balachandran, N. Yadati, A. Kumar, P. Rajagopalan, S. Sanyal, P. Talukdar
  NeurIPS 2018 Workshop on Machine Learning for Molecules and Materials (2018)
  NeurIPS Workshop
  arXiv

Expertise

Education

• 2009 – 2014 · Ann Arbor, MI, USA
  Ph.D. in Mechanical Engineering, University of Michigan

  Thesis: Influence of material structure on thermoelectric properties of atomic scale systems

• 2008 – 2009 · Singapore
  M.Eng. in Mechanical Engineering, National University of Singapore
• 2002 – 2006 · Chennai, India
  B.E. in Mechanical Engineering, Anna University