Skip to content
View nickmartins-lambda's full-sized avatar
  • Joined Sep 9, 2026

Block or report nickmartins-lambda

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nickmartins-lambda/README.md

Nick Martins

Engineer across AI, quantitative finance, and data infrastructure —
with a second life in the physics of sound.

AI / ML Quant Systems Acoustics


I build complete AI-data products end to end — and, crucially, I own the compute that runs them: a private multi-node GPU mesh where models are fine-tuned and served at a fraction of cloud cost, with data that never leaves the box. Ten-plus years spanning applied AI, quantitative markets, data engineering, and full-stack systems — usually as the person who both makes the call and ships the thing.

What I do

  • LLM & agentic systems — orchestration, tool use / MCP, RAG grounding, and an agent-driven fine-tuning factory (dataset → LoRA → serve) on owned GPUs.
  • Quantitative finance — FX-futures strategies, a 28-pair currency-strength model, and real-time market-data pipelines (MT5 → Flask → Redis → live dashboards).
  • Data infrastructure at scale — Python→Go rewrites at 30–50× throughput, residential-proxy acquisition, checkpoint/resume pipelines.
  • Full-stack delivery — React / TypeScript front ends over Go / Python / FastAPI services, shipped to production.

Selected work

  • Magma — a live market-intelligence product (my own idea and IP): defensible, lens-structured numbers over Brazilian markets and trade data, exposed as APIs, dashboards, and a research pipeline.
  • DINO — a data-intelligence platform delivering 4M articles · 26 years · all G20 in 72 hours for $78, outperforming $120k/yr enterprise tools.
  • RRTV — a fully automated financial broadcast network: 13 show slots, zero manual editing per episode, one engineer + an AI toolchain.
  • Reverb Extractor — open-source differentiable Feedback Delay Network (PyTorch), reimplementing recent room-acoustics research. (pinned)

The through-line

I started at Berklee (Electronic Production & Design — essentially programming — with a minor in electronics & acoustics, cum laude), came up through C, DSP and Csound in the lineage of the people who wrote the book on it, and carried that signal-processing instinct forward: music-information retrieval and embedded work, then early machine learning, then — chasing continuous, defensible data — into quantitative finance, and now into games and audio R&D. Different domains, one habit of mind: model the system honestly, then make it fast, cheap, and real.

How I work

  • Own the metal. A private RTX-3090 GPU mesh on a Tailscale network — fine-tuning, VRAM-safe serving, cross-site redundancy. Private models, cost-optimized, no cloud lock-in.
  • Eval-first & honest. My methodology work is citable, reproducible, and candid about where it stops — negative results included. I'd rather show I understand the ideas than hide the gold.
  • Polymath by design. Audio · DSP · acoustics · quantitative markets · AI · data engineering · games — the range is the point.

📚 Research & publications

Work I've authored or provided the data science, analysis & visualization for — each with a permanent, citable DOI (auto-imported by Zotero / Mendeley, indexed across the scholarly graph).

Acoustics & signal processingauthored   DOI  Inverting the Room: Differentiable Feedback Delay Networks for Reverb Extraction and Dereverberation

Economics & political economydata science & visualization · Insight Inteligência

  • DOI  O multilateralismo em xeque — Índice de Prestígio dos Organismos Multilaterais (nº 110)
  • DOI  A "Vida Nua" e a Banalidade da Opressão na "Era da Exceção" (nº 112)
  • DOI  Leis, Loucos e Labirintos: o Manicômio Regulatório (nº 114)

ESG & sustainabilitydata · Anuário Integridade ESG

  • DOI 2024   DOI 2025   DOI 2026

Markets & trade intelligencedata science · Relatório Reservado / RR Magma

  • DOI  China trava preços da celulose e pressiona resultados de Suzano e Klabin
  • DOI  Governo prepara ofensiva para aumentar nacionalização de veículos chineses
  • DOI  Relatório Reservado cria plataforma de inteligência e lança o RR Magma
Stack & tooling

Languages — Go · Python · TypeScript · SQL · (reads Java, C) AI/ML — PyTorch · LLaMA-Factory · LoRA/QLoRA · vLLM · RAG · MoE ensembles · NLP (multilingual) Data & infra — FastAPI · Redis · Postgres/MySQL · Docker · LXD · Kubernetes (k3s) · Tailscale · AWS · GCP Frontend — React · TypeScript · Three.js / WebGL Quant — MT5 / MQL5 · real-time market data · time-series Audio/DSP — Csound · differentiable DSP · room acoustics · voice pipelines (STT→LLM→TTS)

Education & certifications
  • B.A. Music — Berklee College of Music (Electronic Production & Design; minor in Electronics & Acoustics) · GPA 3.78, Cum Laude
  • Australian National University — Introduction to Actuarial Science
  • Harvard University — Statistics & R

Open to remote engagements worldwide. Multilingual — EN · PT · ES.

Popular repositories Loading

  1. nickmartins-lambda nickmartins-lambda Public

    Profile

  2. differentiable-fdn differentiable-fdn Public

    Reverb Extractor — differentiable Feedback Delay Network with learnable delay lines (PyTorch reimpl of Mezza et al. 2024)

    Python