Neil Rathod
I build AI systems by day, ship products by night, and study very old books in between.
About
I'm a full-stack AI engineer at Bank of America. Before that I was a computer scientist at Southwest Research Institute, where I shipped mission-critical signal-processing software in Rust to ~45 machines worldwide, made over-the-horizon radar post-processing 25× faster, and wrote (and won funding for) a research grant targeting a million radar pulses a second with modern C++.
These days the thread running through my work is orchestration — getting many models, machines, and moving parts to behave like one system. My current platform routes tasks across Claude, GPT, Gemini, and self-hosted models through a Thinker–Worker–Verifier loop, grounds its own claims against source documents, and runs on a five-node hybrid cluster I administer myself.
Off the clock: I taught 25+ kids to program over two years and directed UT Dallas's first hardware hackathons. I'm learning Sanskrit and five other languages (one at a time — I've learned my lesson), keyboard and guitar, and I read research papers and very old books for fun.
Certifications
- Professional Machine Learning Engineer (GCP, in progress)
- CAPM (PMI, in progress)
- Signal Exploitation & Geolocation (SwRI)
- Agile Project Management (AgileKB)
Experience
- Building AI systems at enterprise scale — full stack, from model orchestration to production UI.
- Led full-lifecycle development of “FlashingForwarder,” a mission-critical signal-processing component in Rust — field-ready across ~45 machines deployed worldwide in 3 months.
- Engineered multithreaded C++/Python optimizations for Over-the-Horizon Radar post-processing: 25× latency improvement.
- Re-architected FFT interfaces for zero-downtime, real-time data streaming into GPU block processing.
- Designed security protocols integrating YubiKey hardware auth with Linux firewalls against physical and network threat vectors.
- Authored and won funding for an exploratory research grant: cache-aware C++23 architecture (std::mdspan, parallel execution policies) targeting 1M radar pulses/second.
- Mentored 25+ students with personalized curricula: OOP, Git, and API design across Python, Java, and C#.
- Built and deployed 3 student projects on AWS; introduced REST API development for games and web apps in C#/.NET.
- Low-level C++ memory optimization work alongside Unreal Engine 5.
- Built Unity/C# AR training workflows adopted for content reaching 70%+ of employees.
- Improved Unifier database payload performance over REST; shipped a Power BI dashboard surfacing methane-activity drivers across well sites.
Projects
Multi-Agent LLM Orchestration Platform
A multi-provider routing system that dispatches every task to the optimal model — Claude, GPT, Gemini, or self-hosted Nemotron — by capability, latency, and cost, through a Thinker–Worker–Verifier control loop. RAG over a domain corpus (pgvector, HNSW embeddings) with an automated verification stage that grounds claims against sources. Runs on a hybrid 5-node cluster (self-hosted + AWS EC2 + Oracle Ampere) over a Tailscale mesh: PostgreSQL, LiteLLM proxy, Ollama, n8n, Grafana/Prometheus. 3D frontend in React 19 + React-Three-Fiber.
Minkowski — High-Performance Math Library
A custom C++/CUDA math library for physics-simulation and game-engine workloads — GPU-accelerated matrix multiplication and linear algebra, up to 10× over CPU baselines.
Real-Time Object Detection Pipeline
Real-time computer-vision pipeline (YOLOv3) with custom CNNs trained in PyTorch; deep metric learning via dlib outperformed MLP baselines by 40%.
This site
Next.js static export on Cloudflare's edge — a portfolio in daylight, with one or two rooms it doesn't show you.
Research
Accelerating Real-Time Signal Processing Pipelines with Modern C++
Authored and secured funding; designed a cache-aware C++23 architecture (std::mdspan, parallel execution policies) targeting 1M radar pulses/second.
Dialysis Buddy — Human-Robot Interaction whitepaper
Synthesized the clinical literature on fatigue and cognitive decline in dialysis patients; architected an HRI framework for an autonomous medical assistant detecting fatigue markers and executing cognitive-soothing protocols.
Skills
Languages
AI / ML
Infrastructure
Systems & Frontend
Education & Leadership
Education
Leadership
- Chair — IEEE Young Professionals · 2025 — 2026
Led early-career development and congressional outreach programs. - Founder & Director — Comet Hack (IEEE @ UT Dallas) · 2020 — 2022
Founded and ran UT Dallas's first hardware-focused hackathons.
Say hello.
Work, collaboration, or a good question — the inbox is open.