Open to ML / AI Engineer roles abroad · visa sponsorship required

I build AI systems that
work outside the notebook.

AI/ML Engineer with 3+ years at Weichai Power, building production RAG search, LSTM predictive maintenance and agentic LLM workflows for industrial systems. My MSc dissertation at LJMU pairs LLMs with multi-agent reinforcement learning for autonomous driving.

View projects Résumé (PDF) Try the live demo GitHub ↗
  • 66%fewer collisions vs a standard RL baseline (MSc dissertation, 65.6%)
  • 60%higher match accuracy in a production RAG engine
  • ~20%less unplanned downtime from LSTM anomaly detection
  • $1.2M+incremental revenue from LLM-generated sales insights

About

Engineer first, then AI.

I started in mechanical engineering (B.E. with Honours in Electric Vehicles), which means I build for real constraints: noisy sensor data, strict accuracy needs and users who are not ML people.

Since 2023 I have worked at a global diesel-engine and hydraulics OEM, moving from hydraulics sales to technical-commercial and engine application engineering while building production AI systems alongside: RAG search, LSTM predictive maintenance, agentic LLM workflows and cloud forecasting. I like owning the whole path from research idea to something a team uses.

Outside work I write about ML and engineering on my blog, explaining fine-tuning, gears and engines in plain language.

Experience & education

Where I've applied it

May 2023 – Present

AI/ML Engineering · Weichai Power, Pune

Built alongside three progressive roles below, applying GenAI and predictive modelling to engineering and sales operations.

  • Architected a production RAG search engine (Qwen2 0.5B, ChromaDB) matching technical engine specs to customer needs: +60% match accuracy, faster sales-to-engineering handoff.
  • Built LSTM predictive-maintenance models with real-time anomaly detection across 25 industrial sensors: ~20% less unplanned downtime.
  • Designed agentic workflows with LangChain and open-source LLMs (Llama 3.1, Qwen2, Gemma2), with tool use, memory, planning and decision tracing.
  • Developed a vehicle fuel-efficiency model and a job-site fuel-consumption algorithm at 97% accuracy.
  • Built Gemma2 demand-forecasting models on AWS, optimising $0.8M+ in annual procurement.
  • Engineered AI digital-twin simulations linking CATIA / SolidWorks CAD with CAD-Llama; automated invoice and document workflows with NLP (40% faster delivery).
  • LLM-generated sales insights for mobile crushers contributed to $1.2M+ incremental revenue (25% over target).
Jan 2026 – Present

Application Engineer, Diesel Engines · Weichai Power

  • Lead engine selection for OEM inquiries, mapping duty cycle, power and torque to the right platform.
  • Coordinated field failure analysis on a 25-truck dump-truck fleet using DiagSmart and historical data analytics.
  • Deployed a field-validation checklist across construction and industrial applications; built standard Long Block BOMs for WP4.6, WP8 and WP13.
Sep 2024 – Jan 2026

Technical-Commercial Engineer · Weichai Power

  • Owned technical-commercial evaluation of hydraulic systems; authored Technical Agreements with OEMs.
  • Single point of contact for 4 major OEMs; ran design reviews to optimise engine-compartment layouts.
May 2023 – Aug 2024

Sales Engineer, Hydraulics · Weichai Power

Managed the full sales cycle for hydraulic pumps, motors and valves, from consultation to close.

Expected 2027

MSc, Machine Learning & AI · Liverpool John Moores University, UK

Dissertation: LLM-Enhanced Multi-Agent Reinforcement Learning for Autonomous Driving.

2023 – 2025

PG Diploma, Machine Learning & AI · IIIT Bangalore

First class with distinction.

2020 – 2023

B.E. Mechanical Engineering, Honours in Electric Vehicles · Savitribai Phule Pune University

First class with distinction.

Selected work

Projects & case studies

Click any card for the problem, approach and result.

Live demo

Neural network playground

A small neural network, written from scratch in JavaScript, training live in your browser. Change the architecture and watch the decision boundary form.

Epoch 0Loss –Accuracy –

Stack: 2 inputs → dense layers → sigmoid output, binary cross-entropy, mini-batch SGD. No libraries.

Toolbox

Skills

Applied AI

RAG architectures · Agentic AI & multi-agent RL · LLM orchestration (LangChain, LangGraph) · Llama 3.1, Qwen2, Gemma2, GPT-4 · fine-tuning & quantisation · NLP · computer vision

ML engineering

Python · SQL · PyTorch · TensorFlow · Scikit-learn · LSTM / CNN / Conv3D / ConvLSTM · ChromaDB & vector search

Cloud & MLOps

AWS (SageMaker, Lambda) · GCP · Azure ML · Docker · Kubernetes

Industrial domain

Predictive maintenance · digital twins · sensor-fusion anomaly detection · hydraulics · diesel engine application engineering · CATIA V5 · SolidWorks · PLM

Certifications

Google Startup School: Prompt to Prototype · SOLID, Docker & Kubernetes (Scaler) · Power BI (Alison) · Python & SQL (HackerRank) · Eicher Trucks drivetrain integration

Writing

Latest from the blog

Practical guides on ML, AI and engineering at mlaiinsightshub.blog.

For the curious

Ask my terminal

anish@portfolio: ~
$

Contact

Let's talk

Hiring for an ML, LLM or applied-AI role, or want to collaborate? I am looking to relocate abroad and will need visa sponsorship, and I reply quickly.

anishrane2000@gmail.com Résumé (PDF) LinkedIn ↗ GitHub ↗

Based in Pune, India · open to relocation abroad