Mantosh Panigrahi
+91-700******* — ****************@*****.***
linkedin.com/in/mantosh-panigrahi — github.com/Mantosh-1996 — huggingface.co/Mantosh12 Professional Summary
Senior Data Scientist with 7+ years of experience delivering high-impact ML and GenAI systems at scale. Spe- cialized in LLMs, RAG architectures, and multimodal AI. Proven track record of reducing operational costs, automating workflows, and improving business KPIs through production-grade deployments. Skills
Programming: Python
Machine Learning & AI: PyTorch, Transformers, CNN, ANN, Word2Vec LLM / Agentic AI: LangGraph, LangChain, Agentic workflows, Text-to-SQL, RAG, PEFT/LoRA, Prompt Engineering, Hugging Face
Data & Analytics: Pandas, NumPy, MySQL, Redshift, EDA, Cube Visualization: Power BI, Google Data Studio, Redash Cloud & Tools: AWS (S3, SageMaker), Docker, MLOps
Other: Web Scraping, Selenium, Pytesseract, Whisper Professional Experience
Senior Data Scientist / Deep Learning Engineer Nov 2022 – Present Analytica Data Labs, Bangalore
• Built an autonomous LLM analyst agent on LangGraph that turns business questions into structured analysis plans, generates and self-evaluates SQL in a bounded retry loop, and writes evidence-grounded data stories, answering 100K+ queries/week across 5 enterprise databases
• Engineered a self-learning “Criticiser” memory (Pinecone + MongoDB hybrid retrieval) that promotes fully verified SQL into deterministic Cube semantic-layer definitions and re-tests partial or unproven patterns until confirmed, cutting repeat errors by 75% without retraining
• Architected a hypothesis-driven reasoning loop in which the agent ranks competing hypotheses, splits a SQL budget across them, and pivots on refuted evidence, using confirmed / refuted / inconclusive verdicts persisted to cross-session vector memory
• Designed two-tier result-validation guardrails (deterministic SQLGlot parsing + LLM semantic review) that catch cross-joins, duplicate-join inflation, cardinality collapse, and null-heavy keys before synthesis, elimi- nating 85% of hallucinated analytical answers on the eval set
• Built a production semantic layer on open-source Cube that grounds the agent in governed business metrics, with a rule-level drafter and generator keeping KPI definitions consistent across P&L reporting
• Fine-tuned LLMs with PEFT/LoRA and quantized them to GGUF for low-cost on-prem inference; deployed GPU-accelerated serving with CUDA and Docker
• Built embedding-based ANN ticket routing and a 9,700+ intent self-service AI workflow, automating 75% of manual triage and deflecting 13% of support tickets
• Led a team of 8 engineers delivering ML recommendation systems that cut ticket creation by 40% in 6 months Data Analyst / Machine Learning Engineer Jan 2022 – Nov 2022 DealShare, Bangalore
• Contributed to 10% market share growth for private label brands through data-driven pricing and strategy
• Built fraud detection models reducing FCFR and NRO by 10%
• Optimized inventory and discount strategies to improve vendor cash flow Data Analyst Jun 2021 – Dec 2021
Snapwiz, Bangalore
• Built dashboards to track product usage and retention using Power BI and Data Studio
• Analyzed user behavior to improve engagement and product decisions Analyst Jan 2019 – Jun 2021
Collabera, Bangalore
• Built real-time dashboards for tracking inventory across 50+ SKUs
• Improved supply chain visibility and operational decision-making Junior Associate Jun 2018 – Dec 2018
Lancesoft, Hyderabad
1
Education
PG in Data Science 2022
IIIT Bangalore
B.Tech in Mechanical Engineering 2018
ITER, SOA University
GPA: 7.83 / 10
2