Live from ICC 2025 in Sacramento, we sit down with Bhavnesh Patel, Go-to-Market lead at RL Core, to unpack how reinforcement learning (RL) is moving from labs to real-world OT/SCADA. We cover RL Core’s agent-based approach (observe → recommend → limited control → scale), where it fits alongside Ignition 8.3, OPC UA, and control systems, and why continuous processes (water/wastewater, solar + storage, oil sands upstream) see fast time-to-value—chemical/energy reductions, smarter setpoints, and adaptable control.

What you’ll learn

- RL vs. traditional MPC and why “data → action” beats “model → action” for many sites

- How RL Core deploys safely in stages and builds operator trust

- Where RL is working today: water/wastewater, H₂S scrubbing, solar + batteries, oil sands

- Practical paths for integrators and operators to pilot and scale

Guest: Bhavnesh Patel — RL Core (OT software startup applying RL to industrial processes)

RL Core: https://rlcore.ai/

Opsite Energy (presenting sponsor): https://opsiteenergy.com/

Chapters

00:00 Intro

00:10 Welcome & ICC 2025 vibe

01:06 Who is Bhavnesh & RL Core’s focus

02:36 ICC “Level Up,” Ignition 8.3, “Prove It”

03:08 What RL Core does (reinforcement learning for OT)

04:04 RL vs basic PID vs MPC; adaptability without re-modeling

05:07 Agent learns by doing; data → action

06:00 Where it sits: SCADA/Ignition + OPC UA (read/write)

07:00 Safe rollout: observe → tiny control → recommendations → expand

08:56 Example outcomes (chemical/energy reductions; continuous improvement)

10:22 Use cases in compression fleets & oil & gas

11:58 Early-stage focus, ideal partners, role of SIs

13:36 “Intrinsically safe” mindset and guardrails

16:00 Continuous vs batch; why ROI is stronger in continuous

16:54 Industries seeing impact (water/wastewater, solar + storage, oil sands)

18:40 Solar + battery dispatch with policy/financial constraints

19:42 Readiness checklist & engagement model

20:59 Closing + next steps

Uygar Duzgun / “Fast Life” / courtesy of www.epidemicsound.com