About Me
AI Engineer building multi-agent systems, RAG pipelines, and production-grade LLM applications.
I'm Suman Ghosh, an AI Engineer with a B.Tech in Computer Science. I build autonomous AI systems — multi-agent architectures, RAG pipelines, and production LLM applications that solve real business problems at scale.
My core work involves designing agent orchestration layers with LangGraph and CrewAI, building retrieval-augmented generation systems with FAISS and ChromaDB, and deploying these as containerized services with FastAPI, Docker, and CI/CD pipelines on AWS and Azure.
I've built systems that process 19K+ medical images with 93% accuracy using transfer learning (VGG-16, YOLOv8), deployed real-time voice AI pipelines handling 100+ concurrent sessions, and architected multi-agent research systems that generate comprehensive reports in under 2 minutes.
I ship production-grade code with proper error handling, observability (Prometheus, Grafana), and monitoring. Every system I build is designed for reliability, maintainability, and scale.
Engineering Principles
Building AI systems with transparency, fairness, and accountability. Every pipeline includes guardrails against hallucination, bias, and data leakage.
Designing systems that handle sensitive data responsibly — encryption at rest and in transit, GDPR-aware pipelines, and privacy-first architecture decisions.
Shipping systems that degrade gracefully: circuit breakers, retry logic, monitoring dashboards, and alerting. AI that works at 2 AM without intervention.
Technical Skills
Core AI / ML / LLM Foundations
- Transformer Model
- PyTorch
- TensorFlow
- Hugging Face
- Python
- Machine Learning
- Deep Learning frameworks
- AI Algorithms
- LLM
- Multi-modal models
- VLLM
- SLM
- RLHF
- Reinforcement Learning
- RL-based alignment
- Diffusion Models
- Stable Diffusion / SDXL
- PEFT (LoRA / QLoRA)
- Embeddings
Data Science & Analytics
- Data Analysis
- EDA (Exploratory Data Analysis)
- Feature Engineering
- Data Preprocessing
- Data Visualization
- Statistical Modeling
- Predictive Analytics
- Time Series Analysis
- Credit Risk Scoring
- Predictive Customer Insights
- Propensity Models
Programming & Tools
- Python
- R
- SAS
- SQL
- SQL Server
- PostgreSQL
- NoSQL (MongoDB, Redis)
- RDBMS (Oracle, MySQL, etc.)
- Git
- GitHub
- Jupyter Notebook
- Scripting (Python, Bash)
MLOps & Deployment
- MLOps technologies
- Model deployment & versioning
- CI/CD
- CI/CD on Linux
- Docker
- Kubernetes
- MLflow
- Azure Pipelines
- AWS SageMaker
- Cloud deployment
- Model lifecycle management
- Model serving infrastructure
- Auto-scaling, load balancing, failover
- Infrastructure-as-Code (Terraform, Bicep)
Data Engineering
- Data pipelines
- ETL / ELT pipelines
- PySpark
- Hadoop
- Big Data
- Airflow
- Luigi
- DAG management
- Databricks
- Google BigQuery
- S3
- Data Management
AI Systems & Architecture
- Design AI agents
- Agentic AI architecture
- Multi-agent orchestration
- Tool orchestration
- State management
- Contextual memory
- Long-term memory systems
- Event-driven systems
- Distributed systems
- System design
- System architecture
RAG & Search Systems
- RAG (Retrieval-Augmented Generation)
- RAG pipelines
- RAG architecture
- Vector databases
- Vector search systems
- Semantic retrieval
- Search engines
- Elasticsearch
- Azure AI Search
- Chroma, FAISS
- Pinecone
- Qdrant
Backend & APIs
- Backend development
- REST APIs
- Django REST APIs
- Flask
- API design
- Secure APIs
- Authentication & Authorization
- API Gateways
Security, Compliance & Governance
- Compliance
- Data Governance
- Enterprise Security
- Data Privacy Laws (DPDP, GDPR)
- Secure coding
- Data encryption
- Secret management
Monitoring & Observability
- Model monitoring
- Model drift detection
- Prometheus
- Grafana
- OpenTelemetry
- Datadog
- Observability stack
- Logging & tracing
AI Applications
- NLP tasks
- Text classification
- Sentiment analysis
- Text generation
- Chatbots
- Document processing (OCR + NLP)
- Recommendation systems
- Fraud detection
- Transaction monitoring
- Image processing
- Computer vision
- Object detection (SSD, Faster R-CNN)
- Custom model training
Automation & Workflows
- RPA integration
- Automation scripts
- Intelligent workflows
- Multi-step reasoning
- Query planning
- Autonomous agents
Agentic & Advanced AI Systems
- CrewAI
- LangGraph
- MCP (Model Context Protocol)
- Tool-augmented systems
- Multi-agent workflows
- Prompt engineering
- Prompt + retrieval strategies
- Hallucination guardrails
Optimization & Performance
- Model optimization
- Quantization
- Compression techniques
- Latency optimization
- Cost optimization
- Reliability trade-offs
- Pipeline sanity
- A/B testing
Cloud Platforms
- AWS
- Azure
- GCP
- Cloud infrastructure
Business / Product Layer
- SaaS product development
- Agile methodologies
- Scrum
- Sprint planning, standups
- Rapid iteration loops
- High ownership mindset
Software Engineering
- Clean code
- Scalable systems
- Testable systems
- Maintainable architecture
- Version control
- Integration patterns (retries, timeouts, consistency)
Specialized Topics
- OCR tools (Document Intelligence, etc.)
- OpenCV
- Cognitive services
- Open-source models
- CAN bus, J1939 (edge/industrial systems)
Evaluation & Experimentation
- Offline evaluation
- Online experimentation (A/B testing)
- Iteration loops
- Performance benchmarking
Documentation & Collaboration
- Document models & assumptions
- Technical documentation
- Internal playbooks
- Agile collaboration
Additional Tools & Libraries
- Scikit-learn
- Keras
- SciPy
- Matplotlib
Infrastructure & Reliability
- SRE (Site Reliability Engineering)
- Observability systems
- Monitoring pipelines
- Logging pipelines
Advanced System Capabilities
- Context-aware agents
- Goal-oriented agents
- Persistent memory agents
- LLM-powered workflows
- Tool orchestration systems