Services
I work on problems where the modeling is the hard part — learned policies for physical systems, surrogates for expensive simulations, causal measurement, and LLM systems that have to survive evaluation. Research through to production. Every engagement starts with understanding your problem.
Reinforcement Learning & Embodied AI
Learned policies for physical systems. I write and train policies in simulation and drive them toward hardware — vision-sensory-motor control for dexterous manipulation, with robustness treated as the deliverable rather than an afterthought.
- ✓ RL policy design and training for manipulation and locomotion
- ✓ Vision-sensory-motor policies for dexterous tasks
- ✓ Sim-to-real transfer and domain randomization
- ✓ NVIDIA Isaac Lab, Gazebo, and ROS pipelines
- ✓ Policy robustness evaluation and failure analysis
- ✓ Motion retargeting and imitation from recorded human motion
Scientific ML & Surrogate Modeling
Neural surrogates for simulations that are too expensive to run as often as you need them, and forecasting on physical systems where the observations are sparse and irregular.
- ✓ Surrogate models for numerical solvers and physical simulations
- ✓ Climate and geophysical time-series forecasting
- ✓ Forecasting under sparse and irregular observation
- ✓ Architecture search via Bayesian optimization
- ✓ Multi-GPU and HPC training (Slurm)
- ✓ Benchmarking against dynamical and persistence baselines
Causal Inference & Experimentation
Model metrics tell you the model got better. Causal inference tells you the business did. I build the experimentation and causal machinery that connects the two, and the statistical framework teams actually adopt.
- ✓ A/B testing infrastructure and experiment design
- ✓ Causal-inference systems for measuring algorithmic interventions
- ✓ Quasi-experimental methods where randomization is not possible
- ✓ Evaluation design for research-to-production transitions
- ✓ Statistical frameworks for cross-functional decision-making
Agentic Systems Design & Development
I build LLM-powered agents that plan, reason, use tools, and execute multi-step workflows. Internal tools, research assistants, customer-facing agents. Reliable and observable.
- ✓ Multi-agent orchestration and coordination
- ✓ LLM tool use and API integration
- ✓ Planning and reasoning architectures (ReAct, chain-of-thought)
- ✓ Human-in-the-loop workflows and prompt engineering
- ✓ Agent evaluation, testing, and AgentOps
- ✓ Game-theoretic incentive alignment
- ✓ Reward modeling and evaluation guardrails
AI Product Development
LLM-powered products end-to-end: architecture, backend, frontend, deployment. I ship the whole thing, ready for users.
- ✓ End-to-end GenAI product architecture
- ✓ Rapid prototyping and MVP development
- ✓ LLM integration and prompt engineering
- ✓ API design, NLP pipelines, and third-party integrations
- ✓ Production deployment and scaling
MLOps & AIOps
Infrastructure to train, fine-tune, deploy, and monitor ML models and LLMs at scale. Reproducibility and observability included.
- ✓ CI/CD for ML and LLMOps pipelines
- ✓ Model fine-tuning, versioning, and registry
- ✓ Automated retraining and drift detection
- ✓ Inference optimization and GPU cost management
- ✓ Infrastructure as code for ML
Search & Retrieval Systems
Production search engines serving millions of users, and RAG (Retrieval-Augmented Generation) pipelines that ground LLMs in your data. I've done both.
- ✓ Production search systems at scale
- ✓ Search relevance tuning and A/B testing
- ✓ RAG pipeline architecture and optimization
- ✓ Vector database design and embedding model selection
- ✓ Hybrid search (semantic + keyword) and reranking
- ✓ NLP-powered document ingestion and chunking
Computer Vision
Custom deep learning vision models for your specific use case. Full lifecycle: data strategy, annotation, training, optimization, edge deployment.
- ✓ Object detection and instance segmentation
- ✓ Image classification and visual search
- ✓ Video analysis and tracking
- ✓ Deep learning model optimization and edge deployment
- ✓ Synthetic data generation
AI Strategy & Technical Consulting
GenAI strategy, vendor evaluation, architecture review, governance. A decade of shipping ML systems across industries.
- ✓ GenAI strategy and AI readiness assessment
- ✓ LLM evaluation, vendor and model selection
- ✓ Enterprise AI architecture design and review
- ✓ AI governance and risk frameworks
- ✓ Proof-of-concept development and team capability building
Let's discuss your project
Book a free 30-minute consultation. I'll listen to your challenge and share honest feedback on whether and how AI can help.
Book a Free Call