Portrait of Venu Gopal Kadamba in a navy suit, outdoors

Hello, I'm

Venu Gopal Kadamba

Staff SWE - AI/ML @ Palo Alto Networks · MS Data Science @ NYU · ML systems · NLP · LLM engineering

I build production-ready ML pipelines spanning transformer models, document AI, RAG, quantization, and AWS deployments. I was founding engineer at Lending Katalyst, building ML systems for AI-powered property title intelligence.

GitHub LinkedIn View projects

About

Background and focus

I'm a data scientist and ML engineer working at the intersection of NLP, structured document understanding, and scalable inference. I have shipped models end to end: research and synthetic data generation, ONNX and OpenVINO optimization, SageMaker deployment, and Kafka-backed serving.

I was Founding Engineer at Lending Katalyst (past entrepreneurship), where I built ML systems for legal and property-title documents (BERT, OCR, CV layouts, AWS microservices). Since then I have contributed enterprise parsing at Perfios (CNN-BiLSTM-CRF, LLM workflows), forecasting and extraction at Crowe, and internships across NLP, CV, and educational content.

Experience

  1. Staff Software Engineer

    Palo Alto Networks, Santa Clara, California
    • Building scalable, secure agents to enhance post-sale workflows and minimize churn.

  2. Data Science Intern

    Intellistack, NYC
    • Transformer-based table understanding; synthetic data for column classification (Higher-Ed and HR); RoBERTa and DistilRoBERTa near 98% F1 in production.
    • Serverless ML inference with ONNX, OpenVINO, and SageMaker; roughly 95% cost reduction (about $50K yearly savings).
    • LLM agents with LangChain and LangGraph plus Claude via AWS Bedrock; semantic retrieval with MiniLM embeddings; Kafka and Avro async pipelines.

  3. Graduate Assistant, Data Analyst

    New York University, NYC
    • Tableau dashboards on NYU Athletics survey data (Qualtrics); statistical analysis on 1.5K+ responses.
    • Python and SQL ETL from the University Data Warehouse for student engagement and attrition trends (100K+ records).

  4. Data Scientist

    Perfios Software Solutions, Remote
    • Attention-based CNN Bi-LSTM CRF model for address parsing over a rule-based baseline; about 12% improvement; currently productionizing supporting business logic.
    • Data augmentation pipeline for complex address scenarios to train neural networks for address splitting.

  5. Data Scientist I

    Crowe LLP, Remote
    • Tabular data extraction pipeline using an attention-based image-to-text translation model; about 7% gain over the prior rule-based approach.
    • Feature engineering on a five-year time series dataset for metal demand forecasting; XGBoost and CatBoost ensembles for one- and six-month horizons.

Entrepreneurship

Past venture · Founding engineer, Lending Katalyst

Founding Engineer

Lending Katalyst Pvt Ltd to

The product was an AI-powered property title tracing platform that delivered instant, customizable reports across English and vernacular documents, verifying chain of title, liens, encumbrances, building plans, and property tax.

  • Title-report generation with image processing, Google OCR, YOLO, EfficientNet, and MobileNet.
  • Custom BERT models for entities and relationships; cut title report generation from about ten hours to about one hour.
  • Production readiness with ONNX, microservices, Docker, AWS Lambda, and AWS ECS.
  • Onboarded three clients and opened further commercial conversations.

  1. Data Science Content Creator (freelance)

    CloudyML Pvt Ltd, Remote
    • R&D content for data science case studies and a Data Structures and Algorithms course.
    • Seoul Bike Trip Duration Prediction project; about 5% better than the baseline approach.
    • Published 200+ videos on ML and deep learning for CloudyML and YouTube.

  2. Junior ML Engineer

    Omdena, Remote
    • Democratizing Access to Maternal Care in Sub-Saharan Africa with Machine Learning: one of 50 collaborators across 20+ countries.
    • Curated datasets from Nairobi, Zimbabwe, Bangladesh, and South Korea; pregnancy-risk model at 95.6% recall with risk-based ranking.
    • Team lead for modeling; weekly stakeholder updates.

  3. Data Scientist Intern

    Argoid Analytics Pvt Ltd, Remote
    • End-to-end information extraction from product descriptions with BERT; about +3.5% F1 vs Bi-LSTM CRF baseline.
    • Bi-LSTM CRF NER with Flair on fashion data; about +4.2% F1 vs rule-based baseline.
    • Parallelized browsing-history recommendations (~35% faster).
    • Python module to extract ingestion fields for search.

  4. Junior ML Engineer

    Omdena, Remote
    • Predicting Cardiac Arrest Using AI: analysis on MIMIC-III and eICU via Python and BigQuery.
    • LightGBM prototypes for two- and ten-hour prediction windows.

  5. Machine Learning Research Intern

    Neuratree Technologies Pvt Ltd, Remote
    • Marathi Word Sense Disambiguation app (PyQt5, word2vec, Bi-LSTM POS).
    • Medical document classification and extraction (NLP + CV): 98.8 F1 classification, 98.2 F1 entity extraction.
    • Waste classification with ResNet-50 and XGBoost (96.7% recall); LSTM anomaly detection on satellite series.

  6. Data Science Intern

    Technocolabs Softwares Pvt Ltd, Remote
    • Custom CNN for American Sign Language (98.7% accuracy).
    • Credit card default modeling after exploratory analysis.

Education

Master of Science, Data Science

New York University

Sep 2024 to May 2026

Big Data · Bayesian ML · GPU architecture · High-performance ML

B.Tech, Computer Science and Engineering

GMR Institute of Technology, Rajam, India

Aug 2018 to Jun 2022

CGPA 8.66 / 10. Merit scholarship (Rs. 10k) in the first year.

Projects

Highlights

GPU-accelerated BPE tokenizer

CUDA tokenizer with cuCollections and CCCL; large speedups vs CPU and Hugging Face; Pybind11 bindings; GPT-2 compatible tokenization checks.

CUDA · C++ · Python

Distributed FlashAttention

Scaling FlashAttention on LLaMA 7B and 1B across a 4-GPU cluster; FSDP, gradient checkpointing, bandwidth benchmarks.

PyTorch · HPC · LLMs

Finance research agent

Llama 3 via Groq, LangChain and LangGraph, Yahoo Finance integration for agentic equity research workflows.

LLMs · Agents

LLM persona alignment

NYU Reinforcement Learning (Spring 2026), Team Reward Hunters: end-to-end study of persona-conditioned dialogue on Llama 3.2 Instruct (1B and 3B) with SFT, DPO, and GRPO at 1k/4k/8k scales. Reproducible QLoRA pipeline (SFT to DPO to GRPO), held-out eval, and multi-turn identity-drift analysis; DPO for the best persona-relevance tradeoff; documented GRPO reward hacking at 4k/8k.

PyTorch · TRL · DPO · GRPO · RLHF

LaRa-Home

Multi-modal agents for floor-plan code compliance and violations; LangChain, LangGraph, Gemini; Streamlit PoC.

Vision · Agents

Big Data movie recommendations

Spark, MapReduce, and MinHash-LSH on 33M+ MovieLens ratings; Jaccard clusters and Pearson validation vs baseline.

Spark · Distributed ML

KVG Movie Zone

Movie search with sentiment and personalization; React and Flask API with TMDB; Docker, Firebase, Heroku.

Full-stack · NLP

AgriAI

React and Flask APIs for crop and fertilizer recommendation from soil features (F1 96.4%); XGBoost, Random Forest, KNN.

Full-stack · ML

Research

Technical writing

Articles

Skills and certifications

Languages and platforms

  • C
  • C++
  • Python
  • SQL
  • MongoDB
  • Bash
  • Java
  • Web

ML and NLP

  • PyTorch
  • TensorFlow
  • Hugging Face
  • Transformers
  • RAG
  • Fine-tuning
  • QLoRA
  • TRL
  • DPO
  • GRPO
  • RLHF
  • Quantization
  • vLLM
  • OCR
  • Layout
  • NER
  • CRF
  • CUDA

Systems and cloud

  • AWS
  • Docker
  • Kubernetes
  • Kafka
  • ONNX
  • OpenVINO
  • MLflow
  • W&B
  • Spark
  • Hadoop

Libraries

  • NumPy
  • Pandas
  • Scikit-learn
  • OpenCV
  • Flair
  • SpaCy
  • LangChain
  • LangGraph
  • ChromaDB
  • Flask
  • Tableau

Certifications

Leadership and volunteering

  • President, CodeChef GMRIT Student Chapter (Nov 2021 to Jun 2022): launched the chapter; workshops and contests.
  • Secretary and web lead, ACM GMRIT Student Chapter (Jun 2019 to Jun 2022): chapter website and technical events.
  • Volunteer, NSS GMRIT (Feb 2020 to Jan 2022): blood donation and education outreach.

Let's connect

Open to collaborations on ML systems, document AI, efficient inference, and responsible ML products.

Email GitHub LinkedIn