Himansh Mudigonda himudigonda

TL;DR

Production LLM systems, multi-agent architectures, and cloud pipelines with measurable impact.

Category	Tools and platforms
Languages	Python, C++, Rust, Go, Java, SQL, CUDA, Bash
LLM Training, Finetuning & Alignment	LLaMA family, Mistral, GPT family, Claude, PaLI-3, Qwen-VL, BLIP, CLIP, DINOv2, I-JEPA
Agents & orchestration	LangGraph, LangChain, LangSmith, AutoGen, CrewAI, LlamaIndex, RAG pipelines, Ray (task orchestration), MLFlow
Training & optimization	PyTorch, JAX, TensorFlow, Hugging Face Transformers, LoRA, mixed precision, quantization, PTQ, distillation, Optuna, ONNX, Triton
Vision & CV	Detectron2, MMDetection, timm, torchvision, OpenCV, Albumentations
NLP infra	Hugging Face tokenizers, sentencepiece, spaCy, sentence-transformers
Retrieval & vector DBs	Weaviate, Pinecone, FAISS, Chroma, Redis Vector, OpenSearch
Databases & storage	PostgreSQL, DynamoDB, MongoDB, Redis, S3, Parquet, Delta Lake
Cloud & MLOps	AWS (EKS, Lambda, SageMaker, API Gateway, Bedrock, Step Functions, CloudFormation, IAM, CloudWatch), Azure, GCP BigQuery, Docker, Kubernetes, Terraform, GitHub Actions
Data engineering	Kafka, Kinesis, Spark, Airflow, Beam, Parquet, Arrow
Backend & APIs	FastAPI, gRPC, REST, WebSockets, GraphQL, auth patterns, Rate Limiting, OpenTelemetry tracing
Observability & QA	Prometheus, Grafana, OpenTelemetry, Loki, Sentry, pytest, chaos testing
Collaboration & delivery	System Design, CI/CD pipelines, ADRs, Design Docs, Code Review, Reproducible Infra and Experiments

Product-first: ship experiments that produce measurable user outcomes.
Metrics-driven: latency, cost, retention, and precision guide design tradeoffs.
Test-first: TDD and benchmark suites validate changes before rollouts.
Cross-discipline: integrate systems engineering, ML research, and product telemetry.

Full time backend/ML roles, contractor AI systems work, tech talks, and collaborations on agentic AI and scalable inference.