CommonAI CIC
CommonAI CIC
CommonAI CIC je v Cambridgeu ustanovljeno podjetje v interesu skupnosti in neprofitna članska organizacija, ki si prizadeva za varen in odgovoren razvoj temeljnih tehnologij umetne inteligence. Njen sodelovalni inženirski model povezuje zagonska podjetja na področju umetne inteligence, uveljavljena podjetja, organizacije javnega sektorja in akademske ekipe z namenom souporabe virov, razvoja intelektualne lastnine ter napredka na področju napredne umetne inteligence v okviru vodenih programov, vključno z visoko zanesljivostnimi pobudami in Scaling Inference Lab. Organizacija članom omogoča dostop do zaščitenega okolja Digital Commons, specializiranega inženirskega in raziskovalnega znanja ter stroškovno učinkovite računalniške infrastrukture za vodene odprtokodne in raziskovalne projekte. Zaposlitev v organizaciji CommonAI CIC ponuja priložnost za prispevanje k praktičnemu in sodelovalnemu delu na stičišču umetne inteligence, raziskav in tehnologije v javnem interesu.

Machine Learning Systems Engineer, Inference Benchmarking

Build reproducible benchmarks for LLM serving as part of CommonAI’s UK AI infrastructure research. Evaluate long-context and agentic workloads on physical systems and in simulations.

Opis

  • Create and maintain reproducible LLM serving workloads, configurations, and benchmarking tools
  • Build long-context and agentic workloads with suitable existing software
  • Characterise request patterns, prompt and output lengths, caching, batching, tool use, and request dependencies
  • Set performance measures alongside application quality and service requirements
  • Work with measurement engineers to reproduce workloads on physical systems
  • Partner with simulator engineers to represent workloads while preserving key behaviours and dependencies
  • Assess how changes to workloads and serving configurations affect conclusions
  • Keep benchmarks current as inference technology and workloads change
  • Record datasets, model and software versions, configurations, provenance, and permitted uses
  • Compare workload performance across systems against agreed application quality and service requirements
  • Help engineers and investors understand infrastructure choices for specific applications

Zahteve

  • Experience building or evaluating machine learning systems, inference services, performance benchmarks, or similar workload-driven systems
  • Strong practical software skills for building dependable benchmark tools, analysing results, and debugging application–infrastructure interactions
  • Understanding of LLM inference and interest in how serving choices affect performance and output quality
  • Care in designing reproducible comparisons and interpreting results against stated application requirements
  • Practical experience with AI coding agents and sound judgement in validating generated code, data transformations, and evaluations
  • Experience with vLLM, SGLang, or similar serving software, long-context workloads, agentic applications, trace analysis, or quality evaluation is an advantage
  • Ability to work with LLM serving; other forms of inference may be considered later
  • Relevant postgraduate study is helpful, but experience and work quality matter more than qualifications
  • Must be based in the UK or attend Cambridge as agreed for hybrid work

Ugodnosti

  • Competitive salary package with pension and benefits
  • Professional development opportunities
  • Networking with influential people across technology and academia
  • Substantial tooling budget for frontier models and coding agents
  • Collaborative, supportive workplace
  • Opportunity to make a significant contribution at a growing organisation
  • Vibrant office near Cambridge train station
  • Hybrid work with Cambridge attendance agreed individually
  • Permanent, part-time, and consultancy arrangements available
  • Flexible start date
  • Inclusive and diverse workplace

Sorodna delovna mesta

Aurora Energy Research

Senior AI Software Engineer

Aurora Energy Research

Lead the development of autonomous LLM systems that support energy market decisions at Aurora Energy Research, whose cloud software tracks and models power grids. Manage teams and deliver agentic systems from early user discussions through production.

Odpri