Akhilesh Raj

I’m a full-stack developer with 1+ years of production experience building AI-powered automation systems at a fintech startup. I’m focused on making automation actually work in production—fast, cleanly, and with practical engineering rigor.

At Finovate Global, I shipped an end-to-end AI Call Disposition System used by QA teams daily. I eliminated ~80% of manual QA effort by delivering an evidence-linked audio timeline player that lets reviewers jump to AI-flagged segments and verify classifications in seconds.

On the technical side, I built a pipeline combining React/TypeScript frontend, Flask backend, Deepgram ASR, and LLM classification for 1,000+ loan outcome categories. I implemented speaker diarization with utterance-level timestamps to separate agent/customer speech before classification, and used client-side audio peak normalization with the Web Audio API to improve transcription quality.

I also engineered deployment-ready AI: I deployed a 20B-parameter GPT-based LLM on Hyperstack cloud GPU infrastructure and engineered prompts/context windows for accurate first-pass classification in a single inference round. I maintained enterprise workflows by exporting JSON audit-ready results for downstream BI tooling, while keeping core classification on-premises via LM Studio and Mistral where needed.