Vetted Research Assistants

“Full-stack engineer who built and owned a production real-time computer-vision inference platform at Credence, spanning Next.js App Router/TypeScript frontend with SSE/WebSocket streaming, a Flask backend, and Postgres analytics. Demonstrated measurable performance wins (70% fewer re-renders; latency cut to ~40–50ms) and strong production rigor (durable orchestration, idempotency, observability, AWS EC2 + CI/CD) with tight post-launch UX iteration based on analyst feedback.”

“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 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.”

“Amazon engineer who built a production AI agent platform (Python/AWS Strands on Bedrock) that lets teams create tool-using, multi-agent workflows—e.g., agents that auto-triage and resolve customer support tickets by reading internal documentation and collaborating with a research agent. Previously worked in Deloitte on IAM using Ping Identity/Ping DaVinci orchestration, and applies orchestration thinking plus structured evaluation (LLM-as-judge, surveys, automated tests) to improve agent reliability.”

“Built ToolDreamer, a framework for selecting relevant tools for LLM agents by training a retriever on LLM-generated reasoning traces, and has hands-on experience building multi-agent systems in AutoGen (MAG-V) focused on question generation and tool-trajectory verification. Currently works as an AI-guides supervisor at Penn State, regularly communicating AI concepts to non-technical stakeholders.”

“Robotics software engineer with experience spanning embedded C++ control on microcontrollers and ROS/ROS2 production systems in automotive and marine robotics contexts (Harbinger Motors, Impossible Metals). Has deep hands-on experience debugging real-time image pipelines (DDS/QoS tuning, HIL stress testing) and building large automated test suites (1200+ tests) plus CI/CD (Dockerized Playwright tests on Jenkins).”

“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.”

“Backend/AI engineer who built a real-time vector database system for high-frequency financial data using Kafka/Flink on Kubernetes, achieving sub-100ms similarity search at 10k+ concurrent load and resolving tricky duplication issues with idempotency/versioning. Also shipped an end-to-end LLM-based travel itinerary feature (profiling + prompt workflows + APIs) with a focus on quality consistency and low latency.”

“Built and shipped an end-to-end podcast recommendation system exposed via a Flask API and React UI, explicitly balancing relevance, diversity (MMR), and safety constraints while meeting ~200ms latency targets. Also implemented a production-style RAG/information-extraction pipeline using web retrieval, spaCy NER, and fine-tuned SpanBERT with guardrails and evaluation loops (precision/recall/F1) to tune confidence thresholds and improve reliability.”

“Robotics engineer with a 2024 internship building an end-to-end software stack for an autonomous humanoid robot that follows natural-language audio commands to make coffee and deliver snacks, including perception (OpenCV), mapping, and ROS Navigation. Also contributing to a robotics foundation model effort by building data preprocessing pipelines using GroundingDINO and SAM2, and has multi-robot coordination experience with algorithms designed to handle real-world communication drops.”

“Built and productionized an LLM-based PDF extraction pipeline for Medicaid policy documents by fine-tuning Gemini Flash 2.0 and deploying via Vertex AI, adding validation/guardrails to improve trust and reliability. Also built and scaled a SaaS platform (cnotes) for cable operators and regularly partners with customers and sales teams through interactive demos, rapid iteration, and real-time workflow debugging.”

“AI/ML engineer with production experience building an LLM-powered resume-to-job matching and feedback product using RAG, with a strong focus on latency, hallucination control, and scalable deployment. Experienced orchestrating ML inference and backend services on Kubernetes and applying rigorous evaluation/guardrail practices; also partnered with business/product stakeholders at Walmart to improve an NLP-based supplier support system.”

“Developed a production wildfire smoke detection system where smoke is visually subtle and easily confused with fog/clouds; addressed this with a hybrid CNN+LSTM+ViT model and multimodal weather features to reduce false positives. Experienced running scalable, reproducible ML pipelines on shared GPU infrastructure using Slurm and Kubernetes-style batch jobs with checkpointing, retries, and rigorous error analysis.”

“Robotics software engineer focused on controls and motion planning for autonomous flight systems using ROS 2 (rclcpp), Gazebo/RViz, and BehaviorTree.CPP. Has hands-on real-time control experience (1ms loop rate) and has improved system performance by tracing latency issues and refactoring vision components (singleton camera init). Also built low-latency Ethernet/TCP comms on top of the IgH Ethernet stack and uses digital-twin simulation (Gazebo, MuJoCo; beginner Isaac Sim) to validate algorithms.”

“Internship experience building production-oriented AI systems, including a real-time voice scam/spoof detector (RawNet + AASIST) hardened for noisy audio via aggressive augmentation and Zoom-based noise simulation, evaluated with EER on clean and wild datasets. Also built an LLM-driven UI automation agent using Playwright for apps like Linear/Notion with modular tool design, unit tests, and replayable scripted scenarios, and has AWS Step Functions experience orchestrating Lambda/Cognito workflows.”

“Backend/platform engineer who built an AI RAG system on FastAPI/Postgres/AWS with 10+ microservices, vector search optimization (ANN + two-stage re-ranking), and GitOps-driven CI/CD that cut deploy time from hours to minutes. Also deployed Java identity services on Kubernetes at TSMC for 200K+ users using ArgoCD/Azure Pipelines, and built a reliable real-time IoT pipeline (MQTT/Node/MongoDB) with strong consistency controls.”

“Research-focused AI engineer who built an agentic pipeline to automatically extract Sphinx-based API documentation/changelogs and generate synthetic tasks for a dynamic LLM code benchmark targeting real-world API evolution and deprecations. Experienced with multi-agent orchestration (AutoGen, LangChain, CrewAI) and rigorous evaluation methods, and has prior multi-agent work from a Microsoft Research internship.”

“Chief of Staff at DataDome and Onna (founding-team experience) who built executive operating cadences, planning processes, and finance-backed dashboards to drive alignment across fast-scaling orgs (200+ global). Led major transformations including a single-product to multi-product platform shift (improving retention and enabling new market entry) and a sales-led to product-led (PLG) transition, and managed fundraising from Seed through Series B with outside counsel.”

“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.”

“Infrastructure-leaning engineer who has built real-time ML systems end-to-end: a Jetson-deployed adaptive Whisper ASR service (Flask + WebSockets, React/TS UI) and a high-throughput Postgres schema for live transcription. Also delivered customer-facing AI billing/OCR improvements for a dental startup (Dentite), boosting OCR performance by 38%, and has experience instrumenting open-source ML deployment stacks to add infrastructure visibility.”

“UPenn robotics researcher who architected a real-time autonomous driving decision-making engine, integrating LSTM trajectory prediction with MPC in CARLA and adding conformal prediction to deliver 95% statistical safety guarantees under strict latency constraints. Also built and debugged an autonomous quadrotor stack with ESKF-based 6-DoF tracking and optimized A*/Dijkstra planning to eliminate latency-induced instability, with experience bridging heterogeneous simulation/control systems.”

“Robotics software engineer who led an end-to-end learning-based UAV controller project, addressing oscillation issues through simulation, gain tuning, and a shift to geometric control. Has ROS experience spanning UAV mocap-based perception and an autonomous driving stack (LiDAR, mapping, AMCL, controller), plus real-world distributed ROS communication over WiFi with performance troubleshooting.”

“UCLA RoMeLa research assistant (since Oct 2025) building an embedded control and sensor-data platform for multi-robot coordination in a simulated warehouse. Deep hands-on experience with ROS on NVIDIA Jetson under RTOS constraints, secure MQTT/TLS telemetry, and SLAM performance optimization (including ORB-SLAM3) validated in Gazebo and deployed via Docker/Kubernetes and CI/CD.”