Jeremy Lancaster

Principal Software Engineer · Distributed Systems & Production AI

Principal software engineer with 20+ years building production systems at scale. Distributed systems, API platform, and AI/ML infrastructure depth. Track record of resolving high-stakes performance crises, architecting systems processing 100K+ ops/min, and shipping production AI deployments with proper observability and reliability discipline. Current focus: local/private LLM infrastructure and retrieval systems for regulated and data-sensitive industries.

Independent Consultant — RSA Industries

2026 – Present
Production AI systems for media and nonprofit organizations · Remote
  • Design, build, and operate the platforms in the case studies above: multilingual video pipeline, hybrid semantic search, multi-tenant product delivery across client organizations.
  • Self-hosted inference on owned hardware (DGX Spark, RTX 4090): model selection, quantization, GPU memory budgeting, serving via vLLM and llama.cpp.

Staff Software Engineer — EasyPost

Feb 2024 – Aug 2026
Core carrier integration team · high-volume shipping infrastructure · Remote
  • Staff IC owning shipping infrastructure processing 180K–200K+ requests/minute across UPS, FedEx, and 230+ carrier integrations; operated at Principal-level scope across architecture, performance engineering, and team development.
  • Resolved production-critical UPS rate-limiting crisis: diagnosed multi-fold request amplification in rate-shopping flows, designed hybrid fallback routing and multi-account round-robin, led vendor escalation with UPS engineering leadership to negotiate capacity increases.
  • Delivered 3× throughput and ~70% CPU reduction on a high-volume Python/Flask carrier microservice — WSGI runtime migration, asyncio anti-pattern fixes, signal-based hard timeouts, backpressure-aware circuit breakers, content-hash request deduplication. Adopted as the reference pattern for async carrier services.
  • Led technical design and delivery of Multi-Piece Shipment integration spanning the Rails monolith and Python carrier services — schema design, validation layer, full backward compatibility across API versions.
  • Architected a Kafka-based distributed transcript capture system with per-service isolated topics feeding centralized log search for self-service carrier debugging.
  • Owned the Ruby/Python technical interview loop with calibrated shipping-domain challenges adopted org-wide; mentored mid-level engineers toward Senior; drafted executive-level incident communications during enterprise-impacting incidents.

Co-Founder & CTO — Ravel Data Inc.

Apr 2024 – Mar 2026
AI-powered data intelligence platform · 515,000+ organizations, 180,000+ websites · Remote
  • Designed and built the production technographic pipeline: Python async Playwright crawler (slot isolation, atomic checkpoints, backpressure) → Go signal processor (1,000+ fingerprints) → exporter with dedup and atomic writes; monthly delivery to enterprise partner S3 buckets.
  • Built the v2 ML account matcher — an 18-feature Random Forest at the end of a four-strategy matching cascade — 92% held-out accuracy at 73% faster runtime and 60% less memory than v1.
  • Built LLM-powered data extraction (OpenAI APIs + local Ollama inference) with Pydantic-validated structured outputs; automated ingestion into PostgreSQL, Salesforce, and a JSON data lake.
  • Developed predictive models for vendor-transition analytics and entity sizing (XGBoost, LightGBM); designed hybrid on-prem/cloud infrastructure (Docker Swarm, MicroK8s) and vector search (pgvector, Milvus, Qdrant).
  • Implemented Salesforce integration with OAuth PKCE and automated CRM enrichment; prepared platform architecture for technical due diligence in acquisition discussions.

Principal Software Engineer — Truist

Oct 2022 – Dec 2023
Cloud architecture, digital lending · Remote
  • Designed and shipped the State Farm partnership integration on AWS Lambda, API Gateway, and cloud-native patterns — $25MM in additional annual revenue.
  • Led a six-Senior-SA team in the cloud-native redesign of the digital lending product; created and led a companywide Cloud Training Initiative with AWS Technical Trainers.

Senior Software Engineer — HubSpot

Feb 2021 – Oct 2022
Internationalization systems, API infrastructure · Remote
  • Migrated a hardcoded Java translation system to an in-house i18n manager — 60% fewer translation errors, near-zero ongoing maintenance; shipped an object-property metadata endpoint serving 1M+ users daily.

Lead Architect — TrackStreet

Apr 2017 – Feb 2021
Distributed data collection platform · Remote
  • Managed a distributed team of 10 engineers across the US and Mexico; reported to CEO, later VP Engineering.
  • Rewrote the data collection system in Scrapy/Python on cloud-native AWS — 800% faster, 99.87% success rate, 34% lower operational cost; built review crawling across 10,000+ websites in under 4 hours of compute.

Earlier — Branding Brand · Eureka! Ranch · Data Conversions · Digital River

2008 – 2017
Senior engineering roles: mobile e-commerce (Hot Topic, American Eagle, Rite Aid), AWS migration, XML delivery and caching infrastructure. Concurrent independent consulting: 200+ engagements.

LLM-JEPA for Theological Reasoning

2026

Applied Meta's Joint Embedding Predictive Architecture to Reformed catechism corpora. Constructed the TheoBench evaluation set (546 pairs); full paper draft. Novel application of self-supervised representation learning to domain-specific reasoning.

Distributed Speculative Decoding

2026

Designed and tested a 5-machine distributed inference setup spanning a DGX Spark and RTX 4090 workstation — Qwen3-30B-A3B at 87 tok/s via tuned llama.cpp. Technical writeup in progress.

Edwin Agent Runtime (Go)

2025 – Present

Go-based agent runtime for production AI workflows.

Shepherd's Study

2025 – Present

Pastoral research RAG tool over commentaries, public-domain sermons, and theological literature — Mistral and Llama base models with custom fine-tuning and persona-based query expansion.

Languages
Python, Go, Ruby, JavaScript/TypeScript, SQL, C
AI / ML
LLM fine-tuning, RAG and hybrid retrieval, vector search, vLLM, llama.cpp, Ollama, HuggingFace, Pydantic-validated structured outputs, OpenAI/Anthropic APIs, local GPU inference (DGX Spark, RTX 4090), faster-whisper, AWQ/GGUF quantization, eval harnesses
Distributed systems
Kafka, RabbitMQ, gRPC, async patterns, circuit breakers, backpressure-aware design, distributed tracing
Databases & search
PostgreSQL (incl. pgvector), Redis, MySQL, Elasticsearch, MongoDB, Milvus, Qdrant
Infrastructure
Kubernetes (incl. MicroK8s), Docker/Swarm, AWS (Lambda, ECS, SQS, Kinesis, API Gateway), GCP, Azure, Railway, Cloudflare, GitHub Actions
Web
Rails, FastAPI, Flask, Django, Scrapy, Playwright, React
Practices
Performance engineering, production observability, CI/CD, technical mentorship, incident response