Ben Sima

ben@bensima.com | 330.809.1699 | linkedin.com/in/bensima

Professional Summary

Staff AI Engineer with 11+ years of experience building and scaling data heavy products. Specializing in LLM inference optimization, agentic systems, developer tooling, and delivering zero-to-one solutions for new customers.

Hard Skills

Languages

  • Clojure
  • Common Lisp
  • Go
  • Haskell
  • Java
  • JavaScript
  • Nix
  • Python
  • Ruby
  • Rust
  • Scheme
  • Shell/Bash
  • SQL
  • TypeScript

Frameworks and Tools

  • Angular
  • Apache Kafka
  • CI/CD
  • DevOps
  • Docker
  • Flask
  • Git
  • Numpy
  • Pandas
  • Ruby on Rails

Employment History

Parasail.io, Staff AI Engineer, 2025/11 - present

  • Sole owner of GPU kernel optimization for a production LLM inference platform: an autonomous agent loop (generator, correctness verifier, Nsight-based profiler) that searches Triton and CUDA kernel space on NVIDIA Blackwell hardware.
  • Built an adversarial correctness gate for machine-generated kernels – held-out shapes, hidden tests, honest TF32 baselines – after establishing that most published kernel "wins" do not survive them.
  • Recovered ~6% serving throughput from a tensor-parallel MoE configuration, measured end-to-end under CUDA-graph capture rather than microbenchmarks.
  • Delivered Parasail Synth, a zero-to-one LLM workflow prototype, from concept to executive review.
  • Designed speculative decoding across disaggregated GPU hosts, moving the draft model off the target node onto Ethernet.
  • Drove prompt-compilation research (DSPy) as a per-token margin lever on dedicated-serverless inference.

Technologies: Python, Triton, CUDA, PyTorch, vLLM, SGLang, Nsight Compute, Nix, Docker.

Groq.com, Staff Software Engineer, Infrastructure, 2019/6 - 2024/10

  • Implemented operations in the high-level ML library groqapi for advanced machine learning workflows.
  • Developed internal and external documentation systems to support both team members and customers.
  • Managed internal services, build tooling, and DevOps processes.
  • Developed hardware-in-the-loop testing to ensure hardware reliability.
  • Built CI/CD pipelines and packaging systems for reliable product delivery to high-stakes customers.
  • Optimized developer workflows, removing bottlenecks and improving cycle time by 25%.

Technologies: Nix, Haskell, Python, bash, Docker, many others.

UE Authority, Senior Software Engineer, 2018 - 2019

  • Developed high-performance data pipelines for UE.co's next-generation platform, enhancing the delivery of car and home insurance sales leads nationwide.
  • Engineered highly concurrent services in Haskell, capable of handling 60,000–70,000 leads per minute.
  • Integrated lead data from external providers and CSV uploads into a unified processing system.
  • Streamlined data scrubbing processes across multiple internal services, improving efficiency and reliability.
  • Delivered processed lead data to front-end applications used by a country-wide network of salespeople.

Technologies: AWS, Haskell data pipeline backend, MySQL and AWS Redshift, RESTful API with Swagger docs

Liaison Technologies, Software Engineer, 2015 - 2018

  • Designed and implemented a distributed, streaming, immutable, event-sourced cloud database for semi-structured enterprise data.
  • Developed AI-based microservices to extract structured intelligence from semi-structured data streams.
  • Enhanced data accessibility by flexibly extracting intelligible information from customers' data banks.
  • Utilized Clojure, Docker containers, and private cloud infrastructure to build and deploy microservices.

Technologies: Clojure and Java, PostgreSQL, ElasticSearch, ArangoDB, Apache Kafka, Apache HBase, REST API with Swagger integration, event sourcing, command query responsibility segregation (CQRS), schema matching

Thermo Fisher, 2017

  • Developed a system to create and run remote hardware tests for an equipment line.
  • Enabled field technicians to collaborate on custom diagnostic tests for individual instruments.
  • Connected instruments across the network for seamless remote access and testing.
  • Facilitated both remote and on-site execution of hardware diagnostics.

Technologies: Typescript & AngularJS UI, Python backend, CouchDB datastore

WholeLatteLove.com, 2015

  • Performed full-stack engineering on a custom e-commerce platform using Ruby on Rails.
  • Designed and implemented an application-wide notification system.
  • Implemented Continuous Integration and Continuous Deployment (CI/CD) processes.
  • Conducted code upgrades and database migrations to enhance system performance.
  • Developed a comprehensive test suite to improve code quality and reliability.

Thinkful, 2014 - 2015

Instructor in data science and full-stack web development with Python for new and intermediate students. Contributed to written curricula.

Freelance developer, 2012 - 2014

Created two ecommerce systems in PHP, a ClojureScript front-end, and e-magazine site for a local Rochester NY publication.

Education

A.B., Philosophy, Rochester Institute of Technology, 2014. Thesis: Wonder and Desire in Romantic Love: Levinas and Irigaray

Performance Reviews

Real quotes from coworkers during anonymous 360 reviews:

Ben works proactively and independently, given even a vague problem description (e.g. "more visibility") he'll come up with something useful.

finds problems where his individual contributions have a very high power-to-weight ratio & implements direct solutions that solve these problems in a timely & efficient manner. has a tendency to cut to the heart of an issue and attack it directly, rather than get stuck in a quagmire of indecision or coordination for issues where the long-term goals are unstable or subject to change.

Establishes what the MVP is and builds it. Avoids bureaucratic overhead.

Ben is open to share his knowledge and help others. He has helped me a couple of times and showed a lot of patience, empathy, and knowledge.