Rajath Rajendra Pai

Rajath Rajendra Pai

Software Engineer

I build developer-experience platforms and automation — backed by a data science & AI background I bring to tooling, analytics, and machine learning.

About

A little about me

I’m a software engineer at Coretura — an automotive software joint venture — where I build the developer-experience platform for engineering teams across partner organizations: GitHub Projects and cross-org backlog automation, AWS CI/CD, observability, and DevSecOps pipelines.

That work builds on an M.Sc. in Data Science & AI, which I bring to analytics, automation, and machine learning. Outside of work I compete on Kaggle across LLM reasoning, AI security, and model optimization — and I swim and cycle. Feel free to reach out if you’d like to connect.

Experience

Where I’ve worked

DevOps / DevSecOps Engineer · Coretura
Mar 2026 — Present
  • Designed and run the cross-org backlog-sync platform — a serverless AWS service (Lambda, DynamoDB, EventBridge) that keeps 6,600+ issues in sync across 5 GitHub organizations and 8 project boards, processing ~19,000 webhook events a month at a 99.99% success rate.
  • Built its automation suite end-to-end: label-gated routing of new issues onto per-org and global Projects, bidirectional field sync across 4 board chains (status, iteration, dates, numbers, text), historical backfills, customer auto-fill, holiday-aware reviewer due-dates, and a daily overdue-label reconciliation sweep.
  • Engineered AWS CI/CD infrastructure — migrated the CI Docker BuildKit cache from Artifactory to S3 (per-branch entries, 14-day auto-expiry), halving devcontainer rebuilds from ~48 to ~24 minutes and unblocking the move to ephemeral runners, and mirrored multi-GB CI images into Amazon ECR (ECR-first pulls) to keep traffic inside AWS and cut external data-transfer costs.
  • Migrated team observability to AWS-hosted Prometheus/Grafana and validated DevSecOps pipelines (Coverity static analysis).
  • Automated partner onboarding end-to-end with a Python tool built on Microsoft Graph and MSAL — bulk Azure AD guest invitations parsed straight from access sheets, with dry-run mode, audit logs, and CSV reports — provisioning access for 80+ partner engineers.
Associate Data Scientist · Volvo Group
Mar 2025 — Mar 2026
  • Built and maintained the Position Management application — one of the most widely used Power Platform solutions in the organization — enabling managers to create positions through a multi-level VP/SVP approval workflow.
  • Designed a Python people-analytics dashboard surfacing turnover, hiring, and diversity metrics; automated certificate generation and built an Employee Flexibility app to support workforce planning.
People Data Analyst · Volvo Group
Aug — Nov 2024
  • Optimized ETL pipelines and Power BI (PowerQuery/DAX) performance, and automated data workflows in Python.
Master’s Thesis Worker — Vehicle Data · Volvo Group
Jan — Jun 2024
  • Built a multivariate time-series model on truck-sensor, weather, and road-type data to predict drivers’ deactivation of ADAS functions, informing system design and usability.
Data Science Intern — People Analytics & Production Logistics · Volvo Group
Summers 2023 & 2024
  • Developed regression models for HR estimation and improved Power BI reporting; applied PySpark/Databricks fuzzy-matching and ML to raise data-matching accuracy on production-logistics datasets within an agile team.
Research Assistant — Vehicle Safety · Chalmers University of Technology
Feb — Jun 2023
  • Analyzed e-scooter ride data for the e-SAFER project to identify usage patterns and safety improvements.

Toolkit

Technologies I work with

Languages
Python · SQL · C++ · Bash
DevOps & Cloud
AWS (Lambda, DynamoDB, EventBridge, S3, ECR, IAM) · Azure AD / Microsoft Graph · Docker · Terraform · GitHub (org admin, Actions, GraphQL API) · Copybara · CI/CD · Prometheus · Grafana · DevSecOps · Jira
ML & Deep Learning
PyTorch · TensorFlow · Hugging Face · ONNX Runtime · MLflow
LLMs & GenAI
RAG · LangChain · LLM fine-tuning & evaluation
Data
Databricks · PySpark · PostgreSQL · SQL Server
BI & Low-code
Power BI · Power Platform

Competitions

On Kaggle

Active competitor across LLM reasoning, AI security, NLP, and model optimization; author of a public starter notebook adopted by the community (105+ upvotes).

AI Agent Security — Multi-Step Tool Attacks

Designed a red-team attack that probes LLM tool-use guardrails (gpt-oss, Gemma) and reverse-engineers the scoring pipeline to surface data-exfiltration and confused-deputy vulnerabilities.

NeuroGolf Championship 2026

Optimized 400 per-task ONNX networks against an exact memory/parameter budget; reproduced the official scorer locally and applied graph-level domain-reduction rewrites plus onnxsim to cut cost while preserving verified correctness.

ARC Prize 2026 (ARC-AGI-2)

Building a hybrid symbolic + neural reasoning solver: a DSL-grounded synthetic-task generator with a LoRA-fine-tuned Qwen model (Unsloth), test-time training, and selection-first inference, fully offline within the 12-hour limit.

MAP — Charting Student Math Misunderstandings

NLP misconception classification with DeBERTa/Qwen + LoRA ensembles; shared a starter notebook adopted by the community (public LB 0.925, 105+ upvotes).

Also competed in NVIDIA Nemotron (LLM SFT/LoRA), Make Data Count ($100k LLM text-mining), CMI (wrist-sensor time-series), and Orbit Wars (RTS agent + self-play simulator).

Projects

Things I’ve built

RAG from Scratch ↗

A retrieval-augmented generation pipeline built from the ground up with Llama 3 and DeepSeek-R1.

Medical Report Generation from Chest X-Rays ↗

CNN (CheXNet) feature extraction with LSTM/RNN decoding to generate radiology reports from frontal and lateral images.

Scattering Parameters Parameterization ↗

Neural networks (Basic, LSTM, GRU) in TensorFlow/Keras predicting input parameters of a human head–antenna system; the GRU model performed best.

AI Learns to Play Flappy Bird ↗

An agent trained via NEAT (neuroevolution of augmenting topologies); game and AI built end-to-end in Python.