Vetted Machine Learning Professionals

“Built an LLM-powered learning assistant (EduQuizPro/EduCrest Pro) that uses RAG over URLs and PDFs to generate quizzes, notes, and explanations for students/professors. Emphasizes production robustness—implemented dependency fallbacks (FAISS/Sentence Transformers/Gradio), CLI-safe mode, and NumPy-based indexing—along with a custom orchestration layer to keep multi-step AI workflows reliable.”

“Built a production GenAI-powered analytics assistant to reduce reliance on data analysts by enabling natural-language Q&A over Databricks/Power BI dashboards, backed by vector search (Pinecone/Milvus) and a Neo4j knowledge graph, including multimodal support via OpenAI Vision. Demonstrates strong real-world LLM reliability engineering with strict RAG, LangGraph multi-step verification, and Guardrails/custom validators, plus broad orchestration and production monitoring experience (Airflow, ADF, Step Functions, Kubernetes, Prometheus/CloudWatch).”

“AI/ML practitioner with experience spanning audio/LLM applications (built "Iota" using Whisper, tiktoken, and a local Ollama-served LLM) and healthcare ML (Facemed.ai; UChicago Radiology). Demonstrates a production-oriented mindset—focus on data/model fit, deterministic field testing, and operational safeguards—and has improved research evaluation workflows via a hash-table-based concurrent model tracking approach.”

“Master’s-level ML researcher with hands-on embodied/edge deployment experience: built a Google Glass motion-tracking system at Sandia using MobileNetV1 + LSTM trained in TensorFlow and deployed via TensorFlow Lite. Has reimplemented transformer-based research for a thesis and demonstrated strong judgment adapting quickly when upstream assumptions changed, and stays current through active reading groups and a JEPA collaboration.”

“ML/NLP engineer currently working in NYC on a system that connects complex unstructured data sources to deliver personalized insights, using embeddings + vector DB retrieval and a RAG architecture (LangChain, Pinecone/OpenSearch). Strong focus on production constraints—especially low-latency retrieval—using FAISS/ANN, PCA, index partitioning, and Redis caching, plus PEFT fine-tuning (LoRA/QLoRA) and KPI/SLA-driven promotion to production.”

“Developed an LLM-driven recommendation agent in Azure Databricks to triage oncology patients and trigger second-opinion case creation using medical claims and EHR data. Uses ICD-10/CPT/J-code features in prompts, embeddings + vector DB similarity, and a backtesting framework emphasizing recall to avoid missing clinically relevant cases while supporting business revenue.”

“ML/NLP practitioner with experience building an embedding-based ad matching and search system at Vericast (BERT embeddings + similarity search) to replace a third-party taxonomy approach, evaluated via a human-curated gold standard. Also built a custom NER pipeline at Allstate for auto accident claims calls using a bidirectional LSTM and achieved 90%+ F1, with a strong emphasis on production-grade ML workflows (testing, CI/CD, orchestration, versioning, validation).”

“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.”

“Insurance-tech CTO and repeat founder with 10+ years in insurance startups; was employee #4/CTO at Polly (formerly DealerPolicy) and helped scale it from a PowerPoint to 250 employees while raising $180M+. Currently building and selling AgentCanvas.ai—an extensible AI accelerator platform for large insurance agencies—after coding the product end-to-end and now running demos/POCs with prospective buyers.”

“Former PwC Director with hands-on early-stage venture experience (e.g., BridgeLights, a big-data analytics concept for early fintech) spanning concept creation, platform architecture, and go-to-market experimentation. Strong focus on building scalable, modular data platforms with rigorous governance/compliance (data lineage, quality controls) and supporting technical diligence in investor-aligned environments.”

“Frontend-leaning full-stack engineer who shipped a civic-tech Next.js/TypeScript app (Boston Police Index) with SSR, dynamic routing, and server-side data fetching, then iterated post-launch with UX and performance improvements. At Walmart (Seller Center micro-frontends), drove large-scale React/TypeScript refactors—standardizing state management, improving hook usage, and cutting type errors ~50%—and earned trust to serve as a code reviewer enforcing quality standards.”

“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.”

“JP Morgan engineer and Texas A&M student web developer who has owned production systems end-to-end, including a real-time ML training workflow that improved internal search relevance by 30%. Experienced with AWS cloud migrations and operating containerized services on ECS with CloudWatch+ELK observability, Terraform infra, and Spinnaker CI/CD; also built event-driven pipelines with RabbitMQ and Elasticsearch at 1M+ record scale.”

“Backend/data engineer who has built production Python microservices (FastAPI) and AWS-native platforms for event ingestion and analytics, combining ECS/Fargate + Lambda with CloudFormation-driven environments and strong secrets/IAM practices. Experienced modernizing legacy logic with parallel-run parity validation and safe phased cutovers, and has demonstrated measurable SQL tuning wins (20–30s down to 1–2s) plus incident ownership in Glue/Step Functions ETL pipelines.”

“Backend engineer with Citigroup experience who built and evolved a self-service user provisioning/identity backend, cutting onboarding from 45 minutes to under 2 minutes. Demonstrates strong production-grade integration and reliability practices (isolated integrations, retries, rollback logic, heavy logging) plus secure API development in Python/FastAPI with OAuth scope-based authorization and incremental, low-risk rollout strategies.”

“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.”

“Backend/platform engineer with experience across C3.ai (supply chain demand planning) and Amdocs (telecom), working on large-scale data systems and microservices. Has driven first-time adoption experiments of Snowflake + Spark to handle billion-record workloads, built Jenkins-to-Kubernetes delivery pipelines with Nexus artifact management, and implemented Kafka streaming between microservices with HA and retry/error-handling patterns.”

“At JP Morgan Chase, built and deployed a production LLM-powered RAG knowledge assistant to help fraud investigators and risk analysts quickly navigate regulatory updates and internal policies, reducing investigation delays and compliance risk. Strong focus on secure retrieval (RBAC filtering), reliability (layered testing + observability), and production constraints (latency/SLOs), with Airflow-orchestrated, auditable ML pipelines.”

“LLM/RAG engineer who has built and deployed corporate-scale systems at Novartis and Johnson & Johnson, including a healthcare AI agent that generates day-to-day treatment schedules. Recently handled a high-stakes safety incident (LLM suggesting overdose) by tightening model instructions and validating with ~200 test prompts, and has strong end-to-end data/embedding/vector DB pipeline experience (PySpark, FAISS, Pinecone) plus SME-in-the-loop evaluation (RLHF).”

“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.”

“At Fileread, the candidate built and deployed an LLM-powered legal document classification and retrieval layer for an agentic extraction system that turns unstructured legal PDFs into structured tables with line-level citations. They productionized a RAG-style pipeline (ingestion, embeddings, retrieval, reranking, generation) and report 95%+ F1 across 70+ legal categories, emphasizing rigorous evaluation and close collaboration with legal domain experts for high-stakes precision.”

“AI/ML full-stack practitioner in a small-scale manufacturing/lab operations environment who deployed a production ML system to improve blood cell order fulfillment by predicting yield/success from donor characteristics. Experienced building custom multi-agent orchestration (Python, LangChain/LangGraph, MCP) and balancing reliability, data quality constraints, and token/ROI economics while communicating tradeoffs to VP-level business stakeholders.”

“Early-career (fresher) candidate who built and deployed a production AI medical document chatbot using a RAG architecture (LangChain + Hugging Face LLM + Pinecone) with a Flask backend on AWS EC2 via Docker. Has experience troubleshooting real deployment constraints (model dependencies, disk space, container stability) and setting up continuous-style evaluation with fixed query test sets tracking relevance, latency, and error rate.”