Vetted Model Fine-tuning Professionals

“At PwC, built and productionized an agentic RAG enterprise search assistant over 6M internal documents (8M embeddings), deployed across AWS and GCP. Drove major retrieval gains (72%→92% precision via BM25+dense hybrid with RRF and cross-encoder re-ranking), reduced hallucinations 30%, achieved <2s latency at 50–60K queries/month, and cut support tickets 30%—boosting adoption to 2,500 users by adding source-cited answers.”

“Worked on onboarding a Middle East logistics client processing thousands of invoices/month, building a production-ready pipeline that routes known vendor PDFs to deterministic regex parsers via Tax ID matching and falls back to LlamaParse for unknown layouts. Added financial consistency validation plus human-in-the-loop review and logging/metrics to continuously reduce LLM usage and improve template coverage.”

“Built and deployed a production RAG-based LLM Q&A and summarization platform for internal documents, emphasizing grounded answers with structured prompting and citations to reduce hallucinations. Experienced orchestrating end-to-end LLM workflows with LangChain plus cloud pipelines (Azure ML Pipelines, AWS), and runs iterative evaluation using both metrics (accuracy/hallucination/latency/cost) and real user feedback to drive reliability.”

“ML/LLM practitioner with experience at Truveta building an LLM-based evaluation framework; identified non-overlapping evaluator failure modes and proposed an ensemble approach that enabled scaling training data and drove ~5% performance gains across multiple internal projects. Strong focus on robustness to distribution shift (augmentation/domain adaptation/meta-learning) and production reliability via monitoring, drift detection, and safe fallbacks.”

“FinTech/AI engineer who has shipped an end-to-end discrepancy-detection product for financial managers using Next.js, FastAPI/GraphQL, Pinecone, and AWS (with dev/staging/prod, observability, A/B testing, and documentation). Also built an AI-native “AI Genesis” system with agentic cyclic workflows, routing, and tool use, and has experience modernizing legacy systems via the strangler fig pattern while coordinating with senior stakeholders on a 5G autonomous simulation platform.”

“LLM/agentic-systems engineer with PayPal experience hardening an LLM-powered fraud support assistant from prototype to production, focusing on low-latency distributed architecture, rigorous evaluation/testing, and security/compliance. Comfortable in customer-facing and GTM contexts—runs technical demos/workshops, builds tailored pilots, and aligns sales/CS with engineering to close deals and drive adoption.”

“LLM/agentic workflow builder focused on productionizing document-processing systems. Redesigned pipelines with LangGraph + RAG, schema-aware validation, and eval/monitoring loops; known for fast incident diagnosis (restored accuracy from ~70% to >95% same day). Partners closely with sales and stakeholders to deliver tailored demos and drive adoption (reported +40%).”

“Built a production RAG-based GenAI copilot backend at Aetna using Python/FastAPI, GPT-4, LangChain, and Azure AI Search, deployed on AKS with Prometheus/Grafana observability. Owned the system end-to-end (ingestion through deployment) and improved peak-time reliability by addressing vector search and embedding bottlenecks with Redis caching, index optimization, and async processing, plus added anti-hallucination guardrails via retrieval confidence thresholds.”

“Built and productionized an AI-native, agentic appeals decisioning system for health insurance operations, automating 500k+ scanned appeals/year. Delivered measurable impact by cutting review time from 12–15 minutes to ~3 minutes and auto-resolving ~85% of cases with strong auditability, evaluations, and human-in-the-loop guardrails, deployed as containerized microservices on Azure AKS.”

“AI engineer who built a production RAG-based internal analyst tool at BlackRock, fine-tuning an LLM on proprietary financial data and adding four layers of guardrails (input/retrieval/generation/output) to improve grounding and reduce hallucinations. Implemented a LangChain-based multi-agent orchestration (7 major agents) deployed on AWS ECS, with reliability measured via internal human evaluation, LLM-as-judge, and RLHF/drift monitoring.”

“Backend/ML engineer who built production LLM-enabled systems at PRGX, including an interpretable contract opportunity scoring engine (Bradley-Terry pairwise ranking) that reached 0.82 weighted Spearman agreement with SME auditors and was integrated into workflow. Also built a Duke student advisor chatbot and hardened it for real-world reliability/security with schema-driven tool calling, normalization, and off-domain defenses; led staged production rollouts with shadow testing and achieved 0.90 F1 on a new extraction field before shipping.”

“Worked on Sophia the humanoid robot, building production animation pipelines and enhancing human-robot interaction via perception and behavior orchestration. Experienced in stabilizing noisy perception-driven state transitions and designing smooth, user-centered behavioral flows, collaborating closely with artists, animators, and experience designers to translate creative intent into measurable system behavior.”

“Built and deployed a production GenAI internal support agent at Bank of America (“Ask GPS/AskGPT”) using RAG on Azure, focused on reducing escalations and improving response quality for repetitive knowledge-based queries. Demonstrates strong production LLM engineering: custom LangChain orchestration, retrieval tuning to reduce hallucinations, rigorous offline/online evaluation, and model benchmarking with dynamic routing (e.g., GPT-4 vs Claude).”

“At Evoke Tech, built a production LLM "Testbench" to quickly compare LLMs/embedding models and RAG strategies (semantic, hybrid BM25, re-ranking, HyDE, query expansion) to select optimal architectures for different client needs. Also developed a multi-agent, multimodal (voice/text) RAG system for live catalog retrieval and safe product recommendations using LangGraph/LangChain with LangSmith monitoring, and regularly translated PM/UX goals into concrete agent behaviors via demos and flowcharts.”