Anthony Maio — AI Researcher

I am not an Ivy‑league or FAANG alumni. I’m the scrappy, old‑school engineer whose leadership calls when they want to ship something hard under real constraints: limited budget, missing infra, no support staff, and a deadline that has already passed. Think NCO, not general – I’m in the trenches with the team and we ship. I’m a staff‑level AI systems engineer who treats agents and models as production actors, not toys. I combine state‑of‑the‑art AI work with 20 years of building and operating high‑reliability platforms in regulated domains (fintech, IAM, sportsbook, crypto). I focus on the systems that make AI actually work in production: ** Memory & scaffolding: Mnemos, a scoped, neuroscience‑inspired memory engine (surprisal gating, affect tagging, reconsolidation, sleep‑like consolidation) that fixes “vector DB = memory” for agents; Synthesis, a federated skill‑based capability layer so agents can be extended without becoming unbounded generalists. ** Models & efficiency: Eve‑2 (272M‑parameter MoE family of specialist nano‑agents) and Eve‑3 SABER (0.5B dense model with novel semantics‑anchored architecture), tuned for deterministic, CPU‑friendly deployments; a structure‑aware tokenizer for schema‑heavy JSON (~20% token savings); Slipstream, a universal comms protocol yielding large coordination‑token savings; CoDA‑GQA‑L, a differential attention mechanism that compresses KV cache VRAM using custom Triton kernels. ** Reliability & safety: Argos‑Swarm for adversarial red/blue teaming and epistemic‑divergence measurement across model swarms; hardened agent runtimes with deterministic replay, sandboxed tools, explicit cost/latency/safety budgets, and CI‑gated trust scoring. Before AI: ** DraftKings – Led an 8‑engineer IAM platform team and a greenfield 0 -> 1 crypto exchange, kept login at 99.99% uptime, and cut infra spend by hundreds of thousands per year. ** Broadridge & startups – Built trading/FX platforms for Tier‑1 banks as a credit/risk SME and founding engineer, with Kafka/event‑sourced systems handling thousands of tps under tight latency and audit constraints. Under Making Minds, I ship open‑source tooling, publish research/essays, and help teams doing hard, safety critical things with AI instead of just bolting on a chatbox. I’m looking for fully remote roles where I’m used as a staff‑level technical leader: defining AI/agent architectures, hardening runtimes, and multiplying a team’s ability to ship safe, observable, cost‑controlled AI systems.,

Stackforce AI infers this person is a Fintech and AI systems engineering expert with a focus on high-reliability platforms.

Location: Danbury, Connecticut, United States

Experience: 18 yrs 9 mos