Full-Stack Engineer and Data Scientist with nearly 5 years of experience building production-grade AI systems, backend platforms, and real-time computer vision solutions.
Last updated: June 2026 · Current OSS sprint: correctness, reliability, and cross-platform fixes across ONNX, OpenTelemetry, React Router, Immich, Appwrite, and production web/backend stacks.
Note: GitHub activity/language widgets skew to recent public repos and under-represent private production work.
- Build and deploy end-to-end AI systems across data, model, backend, and product layers.
- Strong in computer vision, multimodal AI, and low-latency edge inference.
- Production-focused engineer with ownership across architecture, APIs, optimization, and delivery.
- Data Scientist (Applied AI / Computer Vision)
- AI/ML Systems Engineer (MLOps, inference, deployment)
- Backend / Full-Stack Engineer (Python, Go, TypeScript)
- Built an end-to-end AI textile sorting stack at Kapdaa (backend + APIs + UI + CV + hyperspectral models).
- Delivered
>90%garment-type classification across15+classes in real-world sorting conditions. - Achieved hyperspectral fibre classification
F1 > 0.90, including trace blend detection at5-10%. - Optimized Jetson/distributed inference pipelines to
30-40 garments/minand up to~2,100 garments/hour. - Improved targeted enterprise backend execution efficiency by
~30%via Go rewrites and distributed-system improvements.
| Project | What I Delivered | Outcome |
|---|---|---|
| AI Textile Sorting Platform (Kapdaa) | End-to-end stack across APIs, backend, CV + hyperspectral models, and edge deployment | >90% garment classification, F1 > 0.90 on fibre detection, 30-40 garments/min throughput |
| Enterprise Network Platform (Jeavio) | Backend features, distributed-system fixes, focused Go rewrites, deployment/security upgrades | ~30% targeted performance gain, major upgrade delivery, ~80% vulnerability backlog resolved |
| OSS Reliability Contributions | Bug fixes and validation work across CLI tooling, ONNX, telemetry, router, and product platforms | Merged correctness fixes plus active PRs with focused tests, CI follow-through, and maintainer feedback loops |
- onnx/onnx#7821 - hardened ConvTranspose validation for grouped input channels with regression coverage.
- immich-app/immich#28884 - fixed album sharing to skip existing users without failing mixed share requests.
- remix-run/react-router#15066 - passed the configured window through browser URL creation to avoid cross-window mismatches.
- remix-run/react-router#14982 - normalized
assetsBuildDirectoryseparators inreact-router-servefor cross-platform correctness. - kedro-org/kedro#5483 - documented stateful hook patterns for safer plugin and pipeline customization.
- opencv/opencv#28818 - fixed Charuco detector stale output handling by clearing previous results in
detectDiamonds. - Gitlawb/openclaude#691 - prevented
/modelscrash caused by non-string saved model values. - Gitlawb/openclaude#616 - improved startup recovery guidance for provider misconfiguration.
- Gitlawb/openclaude#613 - fixed compact read/edit prefix ambiguity for tab-indented files.
- onnx/onnx#7821 - ConvTranspose validation hardening and regression coverage.
- immich-app/immich#28884 - album sharing idempotency fix reviewed and merged with full CI.
- remix-run/react-router#15066 - browser URL creation fix for configured-window correctness.
- googleworkspace/cli#724 - fixed
--outputfor JSON-wrapped base64 attachment payloads. - googleworkspace/cli#700 - added safe warning behavior for Google Tasks due-time truncation.
- googleworkspace/cli#699 - fixed OAuth proxy URL parameter integrity.
- open-telemetry/opentelemetry-collector-contrib#48064 - added OTLP HTTP endpoint base-path handling in
telemetrygen; following CI and review feedback through completion. - appwrite/appwrite#11774 - fixed sparse-array serialization after VCS pull-request cleanup.
- calcom/cal.diy#28740 - clarified Docker self-hosting URL mismatch troubleshooting and external-contributor CI gate status.
- ResearchHub/researchhub-backend#3280 - targeted bounty notification behavior with regression coverage.
- mlflow/mlflow#22653 - configurable Huey storage URL for Redis-backed server jobs.
Current focus: AI/ML validation, developer tooling reliability, telemetry/data-loss bugs, router/runtime edge cases, and standards-compliant OSS fixes across Python, Go, TypeScript, and C++ ecosystems.
- Software Engineering:
Python,TypeScript/JavaScript,Go,C++,SQL,Bash - Backend and Distributed Systems:
FastAPI,Flask,Django,REST,GraphQL,gRPC, microservices - AI/ML and CV:
PyTorch,TensorFlow,Hugging Face,OpenCV,ViTs,Hyperspectral Imaging,ONNX - Edge and Optimization:
NVIDIA Jetson,TensorRT,DeepStream,OpenVINO, low-latency inference pipelines - Cloud and DevOps:
AWS,Azure,GCP,Docker,Kubernetes,CI/CD,Jenkins,Ansible - Data and Infrastructure:
PostgreSQL,MySQL,MongoDB,Redis,Kafka,Spark, ETL/ELT
- Knee Ligament Tear Assessment - 3D CNN pipeline for ACL tear detection and grading from MRI.
- Emotion-based Music Player - end-to-end full-stack CV application with real-time emotion inference.
- SFO Passenger Survey Analysis - statistical modeling and visualization of traveler experience factors.
- Email:
jeevanpawar5890@gmail.com - LinkedIn: linkedin.com/in/jeevan-pawar
- GitHub: github.com/jeevan6996
Open to opportunities in Data Science, AI/ML Systems, Computer Vision, and Backend Engineering.