Union City, CA · PayPal · Resume · GitHub · LinkedIn · Email
I build agentic AI and data platforms that survive production — LangGraph/MCP agents, enterprise RAG, anomaly detection, and LLMOps — and ship the same rigor in open source (CortexOps, TideVec).
| Role | Senior Data & AI Engineer, PayPal (San Jose) |
| Open source | CortexOps · TideVec · EmbedShift · EAPO |
| Research | Multi-agent metrics · energy-aware prompting · VAE-BiLSTM |
| Next | PhD, CS (Cybersecurity & AI) — Dakota State University, Fall 2026 |
Experience
Senior Data & AI Engineer — PayPal
Feb 2022 – Present · San Jose, CA
Architect and own enterprise AI/ML platforms used for fraud detection, anomaly detection, automation, and knowledge retrieval at PayPal scale.
- Built agentic AI workflows with LangGraph, LangChain, and MCP for multi-step tool use, orchestration, and production agent deployments
- Designed and shipped production RAG with vector search, semantic retrieval, embedding pipelines, and indexing at financial-data scale
- Owned real-time anomaly detection using VAE-BiLSTM and CNN-BiLSTM on Kafka, Spark, and cloud-native infrastructure
- Fine-tuned LLMs with LoRA/QLoRA; improved prompt and model quality for internal automation and knowledge systems
- Established ingestion, schema evolution, governance, lineage, and AI observability frameworks for trustworthy LLMOps
- Mentored engineers on agentic patterns, evaluation, and production ML/LLM practices
Data Engineer — Kforce Inc. (US Bank)
Feb 2020 – Feb 2022 · San Francisco, CA
Data and ML platform engineering for banking analytics, fraud, and risk workloads.
- Optimized Spark/SQL pipelines for ML feature generation, fraud analytics, risk scoring, and operational intelligence
- Built feature-engineering pipelines and event-driven architectures for near real-time analytics
- Implemented data quality, lineage, and validation frameworks that production ML workflows depended on
- Partnered with data science teams to harden models into reliable batch and streaming data products
Data Engineer — Cognizant (US Bank, Barclays)
Mar 2014 – Feb 2020 · San Francisco, CA
Enterprise data engineering across regulated banking and capital-markets environments.
- Built large-scale ETL/analytics on Spark, Hadoop, Python, and SQL for predictive analytics and BI
- Designed ingestion frameworks and analytical models for fraud detection, forecasting, and scoring
- Implemented governance, lineage, auditing, and compliance controls for regulated financial reporting
- Delivered durable pipelines and data products across US Bank and Barclays engagements
Software Engineer — Wipro Technologies
Nov 2010 – Mar 2014 · India
- Developed data warehouse and reporting automation (PL/SQL, Python) for enterprise clients
- Built foundational warehousing and reporting systems that preceded the move into large-scale data platforms
Open source
CortexOps — AI agent reliability
CI layer for agents: golden datasets, tool-call gates, and regression diffs that block bad deploys (LangGraph, CrewAI, AutoGen, …).
TideVec — Temporal vector database
C++20 vector DB with temporal decay and causal graphs — so RAG stops treating every embedding as eternal truth.
Other projects
| Project | What it is |
|---|---|
| EmbedShift | Embedding-space adapters for zero-downtime model migration |
| EAPO | Energy-aware prompt optimization (Bayesian search) |
| ClaimLensAI | Multimodal damage-claim intelligence |
| VAE-BiLSTM pipeline | End-to-end anomaly detection with TFX |
More: github.com/ashishodu2023
Papers
-
A Unified Metrics Framework for Evaluating Coordination Dynamics in LLM-Based Multi-Agent Systems — Zenodo · NeurIPS 2026 (under review)
-
Energy-Aware Prompt Optimization for Large Language Models — TechRxiv · code
-
VAE-BiLSTM Anomaly Detection for Beam Stability in Spallation Neutron Sources — TechRxiv · Jefferson Lab · code
-
TideVec: Production Vector DB with Temporal Decay & Causal Graph — SSRN preprint · code
Skills & education
AI systems: LangGraph · LangChain · MCP · RAG · LoRA/QLoRA · PyTorch · TensorFlow · vector DBs
Platforms: Spark · Kafka · Airflow · Databricks · GCP · AWS · Kubernetes · Docker · BigQuery
Domain: Fintech · fraud & risk · AI observability · governance · research → production
Certifications: GCP Professional Data Engineer · GCP Professional ML Engineer · Azure Data Scientist Associate · Databricks Spark Professional
Education
- PhD, Computer Science (Cybersecurity & AI) — Dakota State University (starting Fall 2026)
- M.S., Data Science & Analytics — Old Dominion University (GPA 3.85)
- B.Tech., Computer Science & Engineering — Kanpur University
Contact
Open to Principal / staff AI & data platform roles (Bay Area / hybrid).