Cracked dev. nerding out.
|
I'm Rohit — a backend-focused software engineer. I care about how systems hold up once they leave the laptop: how services talk to each other, where the bottlenecks hide, what breaks under load, and what it costs to keep things running. Most of my time goes into designing and building production-grade backends rather than chasing surface polish. Things I keep circling back to:
|
|
Going deep, not wide:
- Agentic AI — Agent architecture, agent workflows, tool use, planning, reasoning, multi-agent systems
- LLM Systems — RAG, context engineering, structured outputs, function/tool calling, LLM application design
- Agent Infrastructure — LangGraph, LangSmith, orchestration, state management, observability, evaluation
- Memory Systems — Short-term and long-term memory, conversation state, retrieval, memory architectures
- Backend depth — Advanced Python, FastAPI, backend architecture, API design
- Systems — Distributed systems, microservices, scalability, reliability, production system design
- Security — Authentication, authorization, threat modelling, service hardening
I'm especially focused on understanding how agentic systems actually work under the hood — how agents maintain state, use tools, retrieve context, manage memory, recover from failures, and coordinate multiple steps reliably.
Alongside that, I'm working through the mathematics underneath modern AI instead of treating models as black boxes — linear algebra, calculus, probability, statistics, optimization, neural networks, and gradient descent.
Slow going, but that's the point.
AI-powered astronomical observation platform. Distributed microservices behind an API gateway — FastAPI for the compute-heavy work, Node.js for the application layer, and a real-time recommendation engine on top.
Microservices · FastAPI · Node.js · API Gateway · MongoDB
Multimodal semantic search across image, audio and text. CLIP for visual embeddings, Whisper for transcription, and FAISS for vector search — stitched into one retrieval pipeline.
Semantic Search · Embeddings · Vector Search · FAISS · CLIP · Whisper
My personal portfolio — projects, experiments, technical work, and things I'm currently building.
Web · Projects · Experiments
What I'm actively pushing on right now:
| Area | What that looks like |
|---|---|
| Agentic Architecture | Agent loops, planning, tool use, state graphs, multi-step workflows |
| LangGraph | Stateful agents, graph-based orchestration, checkpoints, interrupts, subgraphs |
| Agent Memory | Short-term state, long-term memory, retrieval, context management |
| LangSmith | Tracing, debugging, evaluation, observability, agent performance |
| Orchestration | Reliable multi-step execution, retries, routing, human-in-the-loop workflows |
| RAG Systems | Retrieval pipelines, hybrid search, reranking, context engineering |
| LLM Applications | Tool calling, structured outputs, agents, LLM-backed features |
| Production Backend | FastAPI, APIs, databases, caching, observability, deployment |
| Distributed Systems | Service decomposition, communication patterns, failure modes |
| Scalability | Load behaviour, caching layers, horizontal scaling |
| Security | Auth flows, authorization, threat modelling, service hardening |
| System & API Design | Contracts, boundaries, reliability, versioning |
| Developer Tooling | Making the loop between idea and running code shorter |
The goal isn't just to build applications with AI.
It's to understand and build the systems around AI:
Models → Tools → Agents → Memory → Orchestration → Observability → Production
That's the rabbit hole I'm currently in.
心臓を捧げよ
Shinzou wo Sasageyo