Tushar Thokdar
Who I Am
I'm the person who takes a 300M parameter foundation model, fine-tunes it on satellite data, and gets it to 4x the baseline performance — then publishes it with docs, inference code, and a reproducible pipeline.
I work across the full ML stack: satellite vision models, Gen AI fine-tuning, RAG pipelines, agentic workflows, and chatbots. From data collection to production deployment on HuggingFace — I don't just train models, I ship them.
Previously built LLM integration frameworks at Unify and satellite computer vision at Godel Space. Won 1st place at an edge AI hackathon deploying optimized models at 4x inference speed.
What I Build
Not just skills on a resume — these are things I've actually shipped and can do for your team.
Satellite AI Models
Fine-tune foundation models (Prithvi, AnySat) on satellite imagery for detection, segmentation, and monitoring.
End-to-End ML Pipelines
Build the full loop: data collection, preprocessing, training, evaluation, and deployment — not just notebooks.
Geospatial Data Engineering
Acquire and process satellite data at scale — multi-sensor fusion, temporal stacking, spectral analysis.
Production Deployment
Ship models with proper docs, inference APIs, containerized environments, and model versioning.
Gen AI & LLM Applications
Build and ship LLM-powered products — from fine-tuning and RAG pipelines to agentic workflows and chatbots.
Backend & Data
Build APIs, process data, and manage databases. Python-native with experience in SQL and real-time systems.
What I've Shipped
Real projects with real results — published models, open-source pipelines, and measurable improvements.
Prithvi-2.0 Flood Detection
Fine-tuned NASA-IBM Prithvi EO-2.0 vision transformer with UPerNet segmentation head on Sen1Floods11 benchmark, achieving SOTA-level flood mapping from Sentinel-2 imagery.
Wildfire Burn Scar Detection Pipeline
End-to-end geospatial ML pipeline from Google Earth Engine data acquisition to Prithvi EO-2.0 fine-tuning, with novel Delta Channel Algorithm for temporal change detection.
AnySat Multi-Sensor Flood Inference
Implemented inference using AnySat multimodal foundation model (CVPR 2025 Highlight) for flood segmentation with SAR-optical fusion from Sentinel-1 and Sentinel-2.
Where I've Worked
AI Engineer (Contract)
Godel Space
Built production computer vision systems for satellite imagery — from data pipelines to model deployment.
- ▸Shipped AnySat multi-modal inference pipeline processing Sentinel-1/2 satellite data
- ▸Built preprocessing pipelines for SAR-optical imagery fusion at scale
- ▸Created reproducible PyTorch training scripts with experiment tracking
- ▸Explored FastAPI for model serving APIs
Delivered production-ready multi-sensor satellite inference system for disaster monitoring.
AI Engineer (Contract)
Unify
Built a modular LLM integration framework for rapid provider switching and A/B testing.
- ▸Integrated GPT-4, Llama, and Claude APIs into prototype applications
- ▸Built content filtering and guardrails reducing harmful outputs
- ▸Created modular API architecture for comparing LLM providers on quality and cost
- ▸Shipped prompt engineering tools and RAG pipelines
Delivered production-ready LLM framework enabling instant provider switching and A/B testing.
Proof I Can Ship
1st Place — AI Startup Hackathon
Edge Runners
Built and deployed an edge AI model in a hackathon setting — achieving 85% of full-model accuracy at 4x the speed. Fine-tuned Phi-3 with quantization and pruning for resource-constrained deployment.
Education
B.Tech Information Technology
Haldia Institute of Technology · 2020-2024
Certifications
Python Programming Essentials
Rice University
A Crash Course in Data Science
Johns Hopkins University
Complete SQL Bootcamp
Udemy
Let's Work Together
I enjoy building things that matter. If you have an interesting problem or a role where I can make an impact, I'd love to hear about it.