{"data":{"jobs":{"edges":[{"node":{"frontmatter":{"title":"Continuous Learner","company":"Verified credentials","location":"Online","range":"Current","url":"https://www.datacamp.com/certificate/DS0025618761485"},"html":""}},{"node":{"frontmatter":{"title":"AI & Backend Lead","company":"SmartOps — TECHgium 2026","location":"Mysore, India","range":"2025 — 2026","url":"https://www.ltts.com/techgium"},"html":"\n- Led the AI/backend work for an AI-assisted incident-resolution prototype: model and pipeline development, training, cloud deployment, and the operating flow behind the proof of concept
\n- Built an Observe → Diagnose → Resolve → Review → Learn workflow with human review before proposed remediation is deployed
\n- Combined spatial-temporal GNN reasoning, Qwen-assisted diagnostics, Amazon EKS, Terraform, Prometheus, CloudWatch, React, and D3
\n- Reached the national finalist stage: one of 34 finalist teams from 62,000+ registrations across 540+ engineering institutes
\n
"}},{"node":{"frontmatter":{"title":"ML Engineer","company":"Cyber Titans — Amazon ML Challenge 2025","location":"Remote","range":"2025","url":"https://www.amazon.science/"},"html":"\n- Led model development, training, submissions, deployment, and AI/backend work with the Cyber Titans team
\n- Used DeBERTa-v3-base with hybrid mean / max / CLS pooling, custom Log-MAE loss, and IQR-based outlier clipping
\n- Finished Rank 83 with a 45.35 SMAPE final submission after disciplined validation and iteration
\n- Placed in the Top 2% of active competitors; the useful lesson was knowing when a leaderboard move was a real improvement
\n
"}},{"node":{"frontmatter":{"title":"Independent Engineer","company":"Private client work","location":"Remote","range":"Ongoing","url":"/archive/"},"html":"\n- Built scoped prototypes across job-signal verification, agricultural sensor telemetry, and encrypted-data modelling research
\n- Translated broad client ideas into small, testable software and ML systems while protecting private names, data, and unpublished implementation details
\n- Kept delivery practical: clarify the problem, make the workflow observable, and leave a usable handoff
\n
"}},{"node":{"frontmatter":{"title":"Open-source Contributor","company":"Keras, SciPy, Matplotlib & Statsmodels","location":"Remote","range":"2025 — Present","url":"https://github.com/rahulrathnavel"},"html":"\n- Landed maintainer-merged fixes in mature Python projects where correctness, reviewability, and regressions matter
\n- Keras #22844: safer IoU metric validation around out-of-bounds class IDs
\n- SciPy #25209, Matplotlib #31707, and Statsmodels #9812: focused fixes for warnings, empty-input crashes, and residual behaviour
\n- Learned to write small changes that another maintainer can quickly inspect, trust, and merge
\n
"}},{"node":{"frontmatter":{"title":"Independent Product Builder","company":"Public prototypes","location":"Chennai, India","range":"2024 — Present","url":"https://github.com/rahulrathnavel"},"html":"\n- Built public prototypes including Omni RAG, Voice Surrogate, Raki, and a controlled blockchain-voting exploration
\n- Worked across retrieval, Android, Kotlin / Jetpack Compose, Flutter, Firebase, on-device interaction, and hardware-connected workflows
\n- Treats prototypes honestly: show the value, document the constraint, and avoid pretending a demo is a production system
\n
"}}]}}}