Zach Siegel

End-to-end engineering, from ETL to UI. Deployed bulk ingestion pipelines for millions of records per day; synced data across transactional and analytical storage; designed customer-facing features.

Generative AI lifecycle & continual learning. Built evaluation and annotation infrastructure to align generated prose with SME preferences, including a custom prompt optimization and fine-tuning framework; a robust named entity resolution service; point-in-time Excel agent harness evaluation.

ML ops. Deployed inference and training (torch) GPU workloads K8s cluster using Nvidia Run:AI. Built custom tooling to manage GPU workloads, load-test, and optimize utilization in inference workloads, including by tracking hyperparameter sweeps of inference options in SGLang and vLLM. Fine-tuned workload-optimized draft models for speculative decoding accelerators to increase draft token acceptance rate by over 50%. Deployed fine-tuned classifier models for both production and training signal.

Quant signal generation. Implemented and built control plane for a quant signal-generation pipeline connecting a deep research agent, stock-picking classifier, and market data for back-testing and trading, which the firm deemd worthy of a 9-figure buy-in. Supported RL pipeline with reward signal annotation tooling and a rephraser pipeline.

Centered software rigor. Maintained high test coverage for backend and frontend Python and Typescript; extended testing paradigms to evaluate nondeterministic AI features. Conducted dozens of technical screens.

Won company-wide hackathon. Placed first of 48 teams with a resume-screening platform that aligns to manager preferences via annotations of synthetic data and automated prompt optimization.

Led widely-used internal software platform (AWS+React+Django) providing: customized genomic design tools; long-running bioinformatics pipelines; scheduling automation and visualization; executive BI analyses; automated handoffs across platforms; integrations with lab robots; IP-related data mining. Onboarded cross-functional contributors, maintained testing suite, documented, and socialized tools.

Automated lab data plumbing by connecting platforms including: lab information database (Benchling), project management software (Smartsheet), and networked lab instruments. Built a QR-based inventory update system. Fully automated several lab instrument data ingest operations.

Developed forecasting models for primate immunogenicity, with a decision-support tool that anticipates spontaneous interruptions to animal availability (a widespread challenge in biotech). Built automated analysis pipeline that delivers increasingly precise recommendations to decision-makers as new assay results are recorded in lab information database. C suite relied on forecasts for (expensive!) contracts.

Taught "Data and Analytics" to students in the MBA programs as part of PhD track teaching requirements.

Created (great) statistics course materials with a JupyterHub/git-based deployment pipeline. Students clicked a "magic link" to access cloud-provisioned, SSO-enabled compute environments.