Open to opportunities

Siddartha A Y

@siddartha

Message

Aspiring Generative AI and Computer Vision engineer building production-ready LLM/RAG systems.

India

Message

What I'm looking for

I’m looking for a role where I can build production-grade LLM/RAG and computer-vision systems—owning end-to-end pipelines, optimizing model performance, and shipping scalable services in cloud or edge environments.

I’m a Computer Science & Engineering student focused on Generative AI, LLM/RAG pipelines, and computer vision—turning research-style models into practical systems. Through my “Summer ML Research Intern” work at Indian Institute of Science (IISc), I helped develop India’s first large-scale open-source urban traffic dataset (UVH-26) and built Vehicle Re-ID models for cross-camera consistency in occlusion-heavy traffic scenes.

At NetAnalytiks, I engineered an LLM-powered recommendation engine using Sentence-Transformer embeddings and deployed a production RAG pipeline (LangChain, FAISS) integrated with Meta LLaMA on AWS. I’ve also built projects like TraceX (hybrid retrieval with BM25 + dense reranking and auditable PostgreSQL/RBAC) and BangaloreNow (GCP cloud-native ingestion/dedup/enrichment) while optimizing edge deployments with model quantization. I lead with hands-on mentorship and open-source collaboration in IEEE and AI/ML club roles.

Experience

Work history, roles, and key accomplishments

Summer ML Research Intern

Indian Institute of Science (IISc)

Jun 2025 - Aug 2025 (2 months)

Contributed to a large-scale open-source urban traffic dataset (UVH-26) by building vehicle tracking, annotation, and preprocessing pipelines. Developed vehicle Re-ID models and fine-tuned multiple detectors (YOLOv11, DAMO-YOLO, RT-DETRv2) to achieve up to 31.5% mAP@50:95 improvement over COCO-trained baselines.

Spatio Temporal Association YOLOv11 RT DETRv2 data preprocessing

Generative AI Intern

NetAnalytiks India

Jun 2024 - Jul 2024 (1 month)

Built an LLM-powered job recommendation engine using Sentence-Transformer embeddings for semantic candidate-role alignment. Deployed a production RAG pipeline (LangChain, FAISS) and fine-tuned OpenAI Whisper for multilingual ASR to reduce manual screening by ~30% in an ATS workflow.

Sentence Transformers Semantic Search LLMs (LangChain Faiss Whisper Embeddings Cosine Similarity Multilingual ASR