DATA & ML SYSTEMS — PARIS, FR

Kushath Nanduri

I build data and ML systems -- from data-readiness tooling to neural audio codecs and reasoning architectures.

STATUS: MSc Data Science, EDC Paris — graduating 08.2026 · Open to data & ML roles from 09.2026

01 / SYSTEMS

Systems Portfolio

forge-prep

data-readiness tooling for LLM fine-tuning

  • audit command emits a 0–100 Forge Readiness Score across six dimensions: volume, quality, dedup, PII, language, format.

KrisCodec

neural audio codec

  • Benchmarked at ~7.4 ms decode latency.

ARIA

recurrent reasoning architecture

  • Frozen GPT-2 Small (124M) + recurrent reasoning core (~20M) + halting controller (~2M) — ~146M total, ~22M trainable.

VOXMAX

shipped voice-scoring product

  • ~12 s of speech yields a VoxScore (0–100) with four sub-dials and one improvement drill.

ZXERO v0

content-access firewall, shipped and sunset

  • Reached first paying customers with $1,247 in live Stripe revenue before sunset — then wound down deliberately after ~5 months, once Cloudflare entered pay-per-crawl and the unit economics no longer justified scaling.

Stickbook

shared scrapbook for group trips

    ZXERO

    AI operating layer, memory + perception + action