Lekhana Kalyan Raj

Built a GPT-4o agentic workflow that scrapes three news sites, drafts content in the magazine’s voice using few-shot prompt engineering, and posts unpublished drafts to WordPress via REST API for editor review—boosting output from ~2–3 hand-written articles/day to 15+ drafts/day. Automated writers’ research with a Python data pipeline triggered by GitHub Actions cron and added a React dashboard to monitor runs. Migrated long-term artist recordings from local Synology to Google Cloud Storage with IAM access controls to reduce monthly storage costs while keeping appropriate access.

Contributed to Neuono, an AI-led fashion label, by training 200+ garment-specific LoRA adapters on Stable Diffusion (via ComfyUI) so the system can tailor outfits. Implemented context-aware styling by building LLM, RAG, and prompt engineering pipelines using FlutterFlow, AlloyDB, and Google Cloud Storage. Reduced the prompt-to-outfit pipeline to under 2 minutes by benchmarking generation providers, integrating fal.ai API, and using a multi-view ComfyUI workflow to keep front/back garment consistency.

Automated diagnosis of BNG and DHCP subscriber login failures by building Python tooling that pulls router state using Junos show commands, parses subscriber logs, and reproduces issues with lab test subscribers—previously handled manually by the subscriber services team.