01
AI & LLMs
9 tools
- LangChain
- LangGraph
- Deep Agents
- Ollama
- MCP
- Agent Workflows
- Tool Calling
- Prompt Engineering
- Context Engineering
The LeadAgentic SystemsRAGKnowledge Infrastructure
AI/LLM engineer
building agentic systems that survive
production.
By Sajid MiyaPatiala, IndiaOriginally NepalOpen to roles
AI/LLM engineer focused on agentic systems, RAG, and knowledge infrastructure. Three shipped full-stack AI products and independent research on transformer-based pronunciation assessment.
Stack: Python, FastAPI, the LangChain / LangGraph ecosystem. I have shipped RAG pipelines over FAISS, ChromaDB, and Pinecone, and wired enough APIs to know that the integration is the product.
0
AI products shipped, end-to-end
0.0 yr
Building agentic systems in production
0.000
Sentence fluency Pearson (HiPPA)
- Agentic systems
- Retrieval-augmented generation
- Tool-using LLMs
- GraphRAG
- Speech assessment
- Multi-tenant AI
- Local-first inference
- Production RAG
- FastAPI + LangGraph
- Vector search
- Memory-aware agents
- Production deployments
- Agentic systems
- Retrieval-augmented generation
- Tool-using LLMs
- GraphRAG
- Speech assessment
- Multi-tenant AI
- Local-first inference
- Production RAG
- FastAPI + LangGraph
- Vector search
- Memory-aware agents
- Production deployments
Currently · 2026
Building
§01
About
The short version.
I build AI systems that survive contact with production. Most of my work is around agentic systems, retrieval, and the boring infrastructure that makes LLMs usable — versioning, permissions, evaluation, observability.
Stack: Python, FastAPI, the LangChain / LangGraph ecosystem. I have shipped RAG pipelines over FAISS, ChromaDB, and Pinecone, and wired enough APIs to know that the integration is the product.
I write runbooks before code. If a system cannot be debugged at 2am, it does not ship. If a model change cannot be evaluated, it does not deploy.
“I write runbooks before code. If a system cannot be debugged at 2am, it does not ship.”
§02
Capabilities
What I do.
The verbs come first. Tools are how I get there.
02
Retrieval & Knowledge
8 tools
- Agentic RAG
- GraphRAG
- Self-RAG
- Corrective RAG
- RAPTOR
- LightRAG
- HippoRAG
- Vector Search
03
Languages & Backend
8 tools
- Python
- C++
- C
- JavaScript
- SQL
- FastAPI
- Flask
- REST APIs
04
ML & Data
13 tools
- Scikit-learn
- Pandas
- NumPy
- TensorFlow
- PyTorch
- NLP
- PostgreSQL
- Redis
- FAISS
- ChromaDB
- Pinecone
- Docker
- AWS
The shortlist
Four pillars · primary stack
Stack01
Python · FastAPI · LangGraph
Vectors02
FAISS · Chroma · Pinecone
Auth & Data03
PostgreSQL · Redis · JWT
Deploy04
Docker · AWS · Linux
§03
Field Record
What I've shipped.
Three full-stack AI products. Reverse chronological.
01
2026
Independent
NexHire
Creator · AI-Native Hiring Platform
End-to-end recruitment platform — job posting through offer — with LLM-driven screening, matching, scheduling, and email.
- 4 third-party API integrations
- Subscription billing live
- Multi-tenant data isolation
- 01Automated resume screening, candidate matching, interview scheduling, and email generation using LLMs, RAG, and agentic workflows.
- 02Integrated Google Calendar, Gmail, Slack, and Stripe for scheduling, communication, team notifications, and subscription billing.
- 03Stack: React, FastAPI, PostgreSQL, Redis, Docker, JWT.
02
2026
Independent
SynapseLearn
Creator · Conversational AI Tutor
Desktop AI learning platform with multi-user auth, RAG over PDFs and web, and persistent memory threads.
- Dual-vector retrieval (FAISS + ChromaDB)
- Multi-user concurrent sessions
- Memory-aware threads
- 01Built secure authentication and conversation management for multiple concurrent users.
- 02Designed a RAG pipeline over PDFs and web resources using FAISS + ChromaDB with contextual retrieval.
- 03Shipped memory-aware chat — persistent threads, session tracking, conversation merging — plus an admin dashboard.
03
2025
Independent
Portfolio Agent
Creator · Local-First AI Assistant
Privacy-first AI assistant. Zero external API dependency. Fully local multi-modal.
- Fully local — zero cloud calls
- Vision + tool calling
- Persistent long-term memory
- 01Built with FastAPI, LangGraph, Ollama, and SQLite — real-time streaming chat with no cloud calls.
- 02Designed persistent SQLite + JSON long-term memory for coherent context across sessions.
- 03Added vision input, tool calling, and dynamic model orchestration for fully local multi-modal deployment.
FastAPILangGraphOllamaSQLite
§04
Research
HiPPA.
Hierarchical multi-task transformer for pronunciation assessment on SpeechOcean762 — extending the benchmark's coverage beyond its phoneme-only baseline.
HiPPA — Hierarchical Multi-Task Transformer for Pronunciation Assessment
Independently built a hierarchical multi-task transformer pipeline for pronunciation assessment on SpeechOcean762, extending the benchmark's coverage beyond its phoneme-only baseline.
Established transformer blocks (WavLM, cross-attention, multi-task learning, CTC) applied to a benchmark whose official baseline only covers phoneme-level scoring.
Read paper (PDF)Headline metric
0.000
Sentence fluency Pearson
+67%relative to the baseline(0.450 → 0.754)
#MetricPearson
02Sentence prosody Pearson0.737
03Sentence total Pearson0.697
The trade-off: a small drop on phoneme-level scoring (0.450 → 0.396), in exchange for three new metrics the benchmark did not previously support — word total, sentence prosody, sentence fluency.
§05
Credentials
Schools & scholarships.
The academic record behind the work.
Academic
01
2023 — 2027
Thapar Institute of Engineering and Technology
B.E. Computer Science and Engineering
Score8.68 / 10 CGPAPatiala, Punjab02
2021 — 2023
St. Xavier's College, Maitighar
School Leaving Certificate (SLC)
Score3.66 / 4 GPAKathmandu, Nepal03
2021
The Old Capital Secondary School
Secondary Education Examination (SEE)
Score3.95 / 4 GPARaniban, Gorkha, Nepal
Awards
COMPEX Scholarship — Government of India (EdCIL)
2023 — Present
Competitively awarded merit scholarship for Nepalese students studying in India. Covers full tuition and hostel at Thapar.