Vetted Research Assistants in Remote

“Built and productionized enterprise LLM/RAG systems, including a Boeing internal solution that gave 400+ program managers conversational access to 1M+ rows of schedule data, with strong emphasis on governance, reliability, and reducing hallucinations in tabular domains. Also has experience running developer-focused workshops (UC Berkeley computer architecture) and partnering with customer-facing stakeholders to drive adoption of a compliance-sensitive NLP product (SEC-aligned) at Penserra.”

greg farhadian

Screened

Senior Software Engineer specializing in cloud data platforms and Java microservices

Remote4y exp

“Backend/data engineer with experience building Kafka-driven real-time pipelines that support ML code deployment and downstream integrations. Currently migrating high-throughput mainframe (COBOL/assembly) processing to Java, using Spark/Databricks to preserve performance and employing rigorous A/B testing across dev/pre-prod/prod with years of historical data.”

“Second most active contributor at Daice Labs, owning a production AI-powered software development collaboration platform’s end-to-end execution infrastructure (TypeScript/Next.js backend, Node.js CLI, shared libs). Built the full multi-agent pipeline (planning/codegen/summary), Supabase-backed context assembly and realtime state, Git/GitHub automation, and a provider-agnostic LLM abstraction with strict Zod validation and retries, backed by extensive tests and design specs.”

“AI/ML practitioner who built an Indian Sign Language recognition system (MediaPipe hand keypoints + CNN/RNN) as an accessibility-focused teaching aid, iterating closely with advocacy groups and educators and reaching 92% accuracy. Also has production-scale data migration experience at Saasgenie, using Kubernetes pod parallelization to migrate 1M+ ITSM records with a 5x throughput gain under API rate limits.”

“Backend-heavy full-stack engineer with experience at Intuit (TurboTax Live) and Paytm payments, building and scaling Java/Spring Boot microservices for high-traffic transaction systems. Has hands-on wins improving peak-load performance using Redis/disk caching and Kafka event-driven patterns, plus React/Redux work for web app integration and strong monitoring practices with ELK.”

“Built and deployed an LLM-powered analytics assistant enabling business users to ask questions in plain English and receive validated Spark SQL executed in Databricks, with a Streamlit/Flask UI. Addressed strict client schema-privacy constraints by implementing a RAG strategy and ultimately leveraging AWS Bedrock and fine-tuned reference docs. Also has production ML pipeline experience using Docker + Airflow and AWS (S3/ECS/EC2) for financial classification models.”

“AI engineer with robotics perception experience at Forterra, building and deploying moving-object/obstacle detection models into real-time robot pipelines. Addressed training crashes/latency via sub-batch training and optimizer tuning, and improved debugging using ROS/ROS2 tooling with 3D voxel visualization and color-coded validation.”

Hsi-Chun Wang

Screened

Mid-level Data Scientist specializing in LLM development and scalable ML pipelines

Remote4y exp

“Built and deployed production LLM pipelines for evidence-based scoring in two domains: biomedical literature mining (scoring ~2700 drug compounds vs gene targets/mechanisms) and long-horizon news analytics (35 years of Chinese articles). Emphasizes reliability at scale (retries/checkpointing/validation), rigorous empirical model benchmarking (GPT-4o/mini/5), and translating results into stakeholder-friendly visual narratives.”

“Built and shipped production RAG-based LLM agents automating multi-step document query workflows, emphasizing reliability via monitoring, retries, structured exception handling, and fallback retrieval (alternative embeddings/keyword search). Demonstrated measurable gains (18% latency improvement, 25% retrieval efficiency, 12% precision) and has experience integrating agents with messy tax and transaction data at RSM using validation/cleaning and idempotent design.”