Headline

Data Scientist & ML Engineer | Specializing in Agentic RAG, LLMs & NLP | Python, AWS, Computer Vision | Building Scalable AI Solutions

About

I build end-to-end AI systems that solve real business problems, from raw data to deployed models. As a Data Scientist, I specialize in engineering multi-agentic RAG systems and LLM-powered applications. AI & Machine Learning: Agentic RAG, Dense Semantic Search (Pgvector, Pinecone), Large Language Models. I focus on building highly technical, production-grade solutions with strict cost optimization and hallucination-guarded synthesis. Let's connect at eng.ahmed.hzb@gmail.com.

Location

Mumbai

Work experience

Data Scientist at MEP QS Services, Apr 2025 - Nov 2025 · 8 mos, Mumbai, Maharashtra, Architected and delivered a production-grade 10-layer Agentic RAG System - Tender Intelligence Platform serving 4 departments (Cost Consulting, Client, Vendor, Contractor) across 12 engineering domains, processing BOQ analysis, variation order disputes, vendor compliance checks, and contractor qualification reviews through intelligent LLM-powered agents. Engineered a hybrid retrieval pipeline combining dense semantic vector search (Pgvector / Pinecone, 3072-dimensional embeddings) with BM25 keyword indexing (Elasticsearch) fused via Reciprocal Rank Fusion (RRF), enabling precise retrieval of exact clause codes, drawing references, and rate schedules that pure semantic search misses. Designed and implemented 48 isolated vector namespaces (dept::engineering_domain schema) with 7 context-aware chunking strategies, including semantic-boundary, table-row, and hierarchical-section chunkers, ensuring retrieval precision across heterogeneous document types: BOQs, tender drawings, vendor datasheets, method statements, and commissioning records. Built a 3-tier LLM cost optimisation framework with intelligent model routing (GPT-4o-mini for lookups → GPT-4o for reasoning → Opus for complex multi-doc verification), semantic query caching achieving a ~60% cache hit target, and LLMLingua context compression delivering 3× context window reduction. Implemented a hallucination-guarded LLM synthesis layer with chain-of-thought reasoning, confidence-threshold gating, cross-document conflict resolution, and numbered citations linking responses to source document, page, and clause, ensuring outputs are fully auditable and trustworthy for high-stakes tender decisions. Designed a RLHF-backed continuous improvement loop: user correction signals trigger async re-embedding via MongoDB/Celery, with stale chunk invalidation, immutable GDPR-compliant audit logs, and Grafana dashboards tracking MRR, NDCG, cache hit rates, and per-query LLM spend.

Junior Data Scientist at Insightrix, Mar 2025 - Apr 2025 · 2 mos, Paris, Île-de-France, France, Contributed to exploratory data analysis and predictive modeling pipelines as a Junior Data Scientist in a fast-paced research and insights environment, delivering data-driven outputs within compressed 2-week sprint cycles. Performed end-to-end data wrangling, feature engineering, and statistical analysis using Python (Pandas, NumPy, Scikit-learn) to surface actionable insights from structured datasets. Collaborated cross-functionally with senior data scientists to prototype and validate ML models, accelerating the team's analytical output and shortening the model iteration cycle. Applied NLP techniques and data visualization to interpret survey and textual datasets, translating complex patterns into clear, stakeholder-ready reports. Demonstrated adaptability and rapid onboarding — independently contributing to live projects within the first week, building confidence in agile data workflows and team-based ML development.

Data Scientist at Insightrix.EU, Mar 2025 - Apr 2025 · 2 mos, Performed automation for data cleaning, preprocessing, A/B testing, managed business KPIs, data warehousing and feature preparation to ensure accuracy and reliability of analysis. Conducted exploratory and basic statistical analysis over data to identify trends, patterns, and anomalies, also experiment tracking (ML flows). Prepared clear reports and visualizations using public research data over operations and management data to communicate findings to non-technical stakeholders. Maintained proper documentation of data sources, automate workflows, and analysis outputs in line with company standards. Collaborated with cross-functional teams to understand data requirements and project objectives.

Data Scientist at Deepak Cyber Solutions, Jun 2023 - Jan 2025 · 1 yr 8 mos, Remote

Data Scientist at Deepak Cyber Solutions, Sep 2022 - Jun 2023 · 10 mos, Hybrid

Data Scientist at Deepak Cyber Solutions, Feb 2022 - Aug 2022 · 7 mos, Hybrid

Education

Bachelor of Technology - BTech, Computer Science & Engineering with Specialization in Artificial Intelligence, Jaipur Engineering College

Higher Secondary Certificate, Biology/Biological Sciences, General, ANNADA COLLEGE, HAZARIBAG