About
Arpan Nookala
AI Engineer
Right now I'm working two AI engineering jobs. At Parcyl.ai I built ScoutGPT — an agent that answers natural language questions over 200M+ rows of geospatial data using LangChain and DuckDB. At Cloud2 Labs I build the inference stack: vLLM, TGI, RouteLLM routing, full observability with Langfuse and Grafana. I also contribute to Intel's OPEA open source project.
Before that, two years at Rutgers RUCI Lab doing ML research. I got a classification model from 55% to 75% accuracy with deep hurdle models, rewrote some slow R code in Python and got a 6x speedup, and ran Spark pipelines on TB-scale datasets. Published first author at IEEE TENSYMP 2023. Co-authored a Springer paper on deep RL for stock trading.
MS in Statistics and Data Science from Rutgers (GPA 3.8). BTech in Electronics Engineering from SPIT (8.81/10). I care about the whole stack — not because it sounds good, but because the interesting problems are usually at the boundaries between layers.
Outside the Terminal
Outside the keyboard
Formula 1
I watch qualifying like other people watch sport. I look at lap delta charts for fun, have real opinions about floor design, and can tell you exactly why 2022 reshuffled the order.
LEGO
Technic and Creator sets. A Technic gearbox is a mechanical API — each brick has a contract. Snap the wrong ones together and the whole thing collapses. Same deal with ML pipelines.
Beatboxing
Since middle school. It's the one thing that reliably catches people off guard in a first conversation. Still at it.
Minecraft
Trying to build Spa-Francorchamps in Minecraft. It's a work in progress. The elevation changes at Eau Rouge are harder than they look.
Career Arc
The short version
AI Applications Engineer at Cloud2 Labs — enterprise LLM inference infrastructure, RouteLLM routing, OPEA open source contributions
Full Stack AI Engineer at Parcyl.ai — architected ScoutGPT agentic pipeline over 200M+ rows; ML Research Engineer at Rutgers RUCI Lab through Dec 2025
ML Research Engineer at Rutgers RUCI Lab — hurdle models (+20pp accuracy), CTABGAN generators, Spark TB-scale ML pipelines
MS Statistics & Data Science at Rutgers; published first-author paper at IEEE TENSYMP 2023 (Canberra) — deep RL for intelligent traffic control
Co-authored Springer publication (ISMIS 2022) on deep RL for automated stock trading with short selling
Data Scientist at Google via DKSH Smollan — ML pipelines on GCP for predictive analytics across 5,000+ products