Looking for a job

Vishal Banwari

@vishalbanwari

Message

AI engineer at DFKI building LLM evaluation systems; created Grid Copilot’s cited, critic-reviewed reports.

Germany

Message

What I'm looking for

I’m looking for a remote role on UK hours and am open to relocating to India.

At DFKI, I built an evaluation harness for autonomous web agents and LLM systems, measuring task performance, latency, energy, and cost. I also worked on a separate human-in-the-loop training project, building tools to inspect confidence and latent representations.

At ABB, I built an agentic diagnostic system using LLMs, RAG, and tool calling to analyze machine events. I owned retrieval, structured extraction, and evaluation, and delivered a React and FastAPI investigation platform deployed on Azure.

In my independent projects, I built Grid Copilot for root-cause analysis, with a tool-calling agent and critic-reviewed, cited reports. I also created voice-coach, a real-time spoken-practice app, and po-triage, which increased unattended line commits from 56% to 88% with none wrong across eight labelled emails.

Experience

Work history, roles, and key accomplishments

AI Research Intern

ABB

Jan 2026 - Jun 2026 (5 months)

Built an agentic diagnostic system over time-windowed machine events using LLMs, RAG and tool calling; normalized approximately six vendor and firmware log formats into structured records. Owned retrieval, structured extraction and evaluation; implemented generate-validate-retry checks and compared outputs with DeepEval. Delivered a React and FastAPI investigation platform, containerised with Dock

LLM RAG Tool calling DeepEval React fastAPI Docker Azure Azure OpenAI VLLM

Research Assistant

DFKI, German Research Centre for AI

Apr 2024 - Dec 2025 (1 year 8 months)

Built an evaluation harness for testing and benchmarking autonomous web agents and LLM systems, measuring task performance, latency, energy and cost across model scales. Worked on a human-in-the-loop training project using expert annotations, building tools to inspect confidence and latent representations and running GPU training with SLURM.

LLM Evaluation Benchmarking Slurm GPU Training Human In The Loop Latent Representations

Senior Engineer

Rugged Monitoring

Jun 2021 - Feb 2023 (1 year 8 months)

Led analytics architecture for the monitoring platform, covering partial discharge and wider power-grid analytics, and took it through deployments at three international field sites. Ingested device data over Modbus, IEC 61850 and SOAP, mapped signals onto virtual nodes, stored in PostgreSQL, and served to Angular/TypeScript dashboards through .NET Core services. Built an auto-configuration tool t

Modbus IEC 61850 SOAP PostgreSQL .NET Core Redis Angular TypeScript Analytics Power Grids

Software Engineer

Qualitrol

Feb 2017 - Mar 2018 (1 year 1 month)

Developed C++/C# applications and firmware interfaces for industrial monitoring hardware, including deployments at two UAE field sites and ML anomaly-detection modules. Ingested device data over Modbus, IEC 61850, IEC 60870-5, SOAP and OPC and mapped signals with unit and quality onto virtual nodes. Served as Scrum Master for 6+ engineers during the 2018–2021 period.