You will get expert AI consultation on LLM fine-tuning, RAG, and model strategy
Project details
You will get clear, senior guidance on your AI project before you spend weeks building the wrong thing. I spent six years as an Applied Scientist at AWS shipping multimodal AI to production, with published research at CVPR, ICCV, and ECCV and 750+ citations. I will tell you straight whether fine-tuning, RAG, or better data is the right fix for your problem, which open-source or frontier model fits, and what accuracy, cost, and latency to realistically expect. You leave the call with specific, prioritized next steps and a written summary you can act on. Good fit for LLM fine-tuning and post-training, RAG, document and image understanding, synthetic data, and evaluation.
Machine Learning Tools
BERT, ChatGPT, GPT-3, OpenCV, Python, Python Scikit-Learn, PyTorch, scikit-learn, SciPy, TensorFlow, Tesseract OCRWhat's included
| Service Tiers |
Starter
$120
|
Standard
$300
|
Advanced
$500
|
|---|---|---|---|
| Delivery Time | 3 days | 5 days | 7 days |
Number of Revisions | 1 | 2 | 2 |
Model Validation/Testing | - | - | - |
Model Documentation | - | ||
Data Source Connectivity | - | - | - |
Source Code | - | - | - |
Frequently asked questions
About Bhavan
Applied Scientist | LLM Post-Training Multimodal AI & Computer Vision
San Francisco, United States - 4:23 am local time
What I can help with:
- LLM post-training: SFT, DPO, RLHF, and RLVR to make a model reliable at your specific task
- Synthetic data pipelines: generate and verify training data when labels are scarce or expensive
- Multimodal and vision-language models: document understanding, VQA, chart and image reasoning, fine-tuning VLMs
- Evaluation you can trust: custom eval harnesses, LLM-as-judge, catching silent quality regressions
Selected background:
- Shipped multimodal AI in Amazon Textract and Bedrock Data Automation
- Co-first-author, CVPR 2024, on LLM-driven synthetic data (published at CVPR, ICCV, ECCV)
- EB1-B Outstanding Researcher (US government recognized), 2 US patents
- MS Robotics (CMU), physics background for first-principles problem solving
I also work on ML for biology: protein and sequence models, single-cell and multi-omics, and interpretable models.
Tell me what you're building and the failure mode you're hitting, and I'll tell you straight whether fine-tuning, RAG, or better data is the right fix.
Steps for completing your project
After purchasing the project, send requirements so Bhavan can start the project.
Delivery time starts when Bhavan receives requirements from you.
Bhavan works on your project following the steps below.
Revisions may occur after the delivery date.
Review your use case and problem
I read your requirements and any materials you share, and note the key technical decisions and open questions before we meet.
Live consultation call
We work through the technical decisions together: fine-tuning vs RAG vs prompting, model choice, data plan, and realistic accuracy, cost, and latency.