// query

Ask about Souravsing, and watch the pipeline that answers you.

// about

Context window

Souravsing builds AI systems that actually ship — not demos. Over 3+ years at Amdocs, he's designed and deployed production-grade LLM, RAG, and agentic AI pipelines, working with a Forward Deployed Engineer mindset: embedding directly with stakeholders to turn ambiguous business problems into working software.

His stack spans the full generative AI lifecycle — retrieval pipelines and vector search, multi-agent orchestration, prompt engineering, and the unglamorous backend work (Flask/FastAPI microservices, Docker, CI/CD) that turns a working model into a system people can rely on.

50%
manual effort reduced via RAG automation
~35%
inference latency improved
40–45%
factual response quality lifted
3+ yrs
shipping production GenAI systems
// experience

Pipeline log

Software Developer — Generative AI & Agentic Systems
Aug 2022 — Present
Amdocs, Pune, India
  • Architected end-to-end Generative AI pipelines integrating LLMs with RAG to automate data-intensive manual operations, cutting manual effort by 50% across target workflows.
  • Built LLM-based agentic systems using LangChain Agents, custom toolkits, and internal APIs for multi-step reasoning, dynamic task planning, and autonomous tool/API calling.
  • Developed and deployed secure AI microservices in Python (Flask/FastAPI), orchestrating model inference and downstream business logic.
  • Engineered a scalable inference architecture with response caching and parallel request processing, improving latency by ~35% under high traffic.
  • Designed model monitoring and evaluation utilities to track accuracy, reduce hallucinations, and maintain groundedness in production.
  • Acted as technical liaison between engineering and business stakeholders — scoping AI use cases and translating them into deployable solutions in agile sprints.
LangChainFAISSChromaDB FlaskFastAPIPython DockerMySQLOracle
// projects

Indexed documents

DOC_0012023 — 2024
Data Gathering & Query Agent

An end-to-end private RAG pipeline enabling secure semantic search over organizational data, without exposing sensitive content to external services.

  • Autonomous multi-step reasoning agents for dynamic task planning and API calling.
  • LLaMA 3.2 for low-latency inference, plus hybrid retrieval, re-ranking, and custom prompt templates — lifting factual consistency by 40%.
  • Role-aware conversational layer supporting multi-agent collaboration across multi-turn workflows.
LangChainFAISSChromaDBLLaMA 3.2
DOC_002Dec 2025 — Jan 2026
Solr AI Query & Update Studio

Converts plain-English requests into optimized Solr queries and automates multi-system document updates — no hand-written Solr syntax required.

  • Natural-language-to-Solr query engine using a local LLM (Ollama).
  • Multi-system update sync across Unix XML files and Oracle databases from one natural-language command.
  • Companion Chrome/Firefox extension for in-context query and update workflows.
OllamaApache SolrOracleXML
// stack

Embedding space

generative ai & llms
LLMsTransformersHugging FaceOpenAI APILLaMAMistralPrompt Engineering
rag & retrieval
RAGHybrid SearchRe-rankingQuery RewritingSemantic SearchFAISSChromaDBPinecone
agentic systems
Agentic AIMulti-Agent SystemsLangChainLangGraphLlamaIndexAutoGenCrewAI
backend & infra
PythonFlaskFastAPIREST APIsMicroservicesDockerCI/CDGit
data & evaluation
MySQLOraclePL/SQLBLEUROUGECosine SimilarityGroundedness Testing
// contact

Endpoint

$ curl -X GET souravsing.dev/contact
{
  "status": "open_to_new_roles",
  "roles": ["Generative AI Engineer", "Agentic AI Engineer", "Forward Deployed Engineer"],
  "location": "Pune, India",
  "email": "souravsingp@gmail.com"
}