Data Scientist

About 1BUY.AI We are building the

Cognitive Engine for global manufacturing . In a world of fragmented data and brittle supply chains, 1Buy.ai leverages multi-agent AI to automate the discovery, risk-assessment, and procurement of electronic components at global scale. Backed by a founding team that has built, scaled, and exited some of India's most iconic B2B and SaaS unicorns, we are at an inflection point and this role sits at the centre of what we ship next.

About the Role We're looking for a

Data Scientist who writes production code every day , ships models that receive real traffic, and is as comfortable debugging a RAG pipeline as tuning an XGBoost model for pricing signals. This is a

hands-on individual contributor role reporting to the CTO, working closely with the ML Engineer. You are not waiting to be assigned problems — you identify gaps, propose solutions, and ship independently. This is

not a people-management, strategy-deck, or org-building role. We want someone who wants to go deep, stay technical, and own the full ML lifecycle end-to-end in a genuinely complex, high-stakes domain.

1. Classical ML & Predictive Intelligence

• Build and own production models for pricing signals, demand forecasting, lead-time prediction, lifecycle risk, and inventory-age risk using XGBoost, LightGBM, CatBoost, and time-series methods (ARIMA, Prophet, temporal Transformers)
• Own the full ML lifecycle: feature engineering, training, offline evaluation, A/B testing, deployment, drift monitoring, retraining
• Decide when a gradient-boosted model or simple heuristic beats an LLM — optimizing for cost, latency, and explainability
• Build SHAP-based interpretability outputs for trusted, auditable models
• Design and run A/B tests to measure business impact within the first sprint in production

2. GenAI Engineering & RAG Pipelines

• Build RAG pipelines for datasheet extraction, BOM parsing, RFQ drafting, and supplier communication
• Design embedding strategies, chunking approaches, hybrid dense-sparse retrieval (Qdrant, Weaviate, pgvector), and reranking layers
• Integrate frontier model APIs (OpenAI, Anthropic, Gemini) and self-hosted OSS models (Llama 3, Mistral, Qwen via HuggingFace) with documented cost/latency/accuracy trade-offs
• Build version-controlled, systematically evaluated prompt engineering systems
• Implement LLM evaluation pipelines (RAGAS or equivalent) to track retrieval quality and hallucination rates

3. Feature Engineering & Data Pipelines

• Build and maintain feature pipelines using dbt, Pandas, Polars
• Work with Postgres, Redshift, S3 to build reproducible feature stores with clear lineage
• Implement data quality checks and leakage-free cross-validation (temporal splits, group k-fold, nested CV)
• Build and document reusable feature registries

4. MLOps, Experiment Tracking & Monitoring

• Own experiment tracking via self-hosted MLflow
• Containerize inference services (Docker) and deploy on AWS (SageMaker, Lambda, ECS) with clear SLAs
• Implement production monitoring (Evidently AI or equivalent) with automated drift/KPI alerts
• Maintain CI/CD pipelines (GitHub Actions) for model deployment
• Use AI-assisted dev tools (Cursor, GitHub Copilot, Windsurf) as first-class practice

5. Agentic AI & Tool Integration Support

• Build Python tools, function schemas, and data connectors for multi-agent orchestration (LangGraph, CrewAI, AutoGen)
• Design MCP-compatible tool endpoints for model-serving and scoring functions
• Support agentic workflow testing and regression tests for agent-facing ML components
• Stay current on the agentic AI/MCP ecosystem; review LangGraph state graphs and flag failure modes

Must-Have

• Ships code daily; active GitHub; Python as primary language; reviews own PRs
• 6–12 years of hands-on experience in data science, ML engineering, or software engineering with a clear arc toward production AI/ML systems
• At least 3 years owning production ML systems (real traffic, real failures, retrained and monitored over time — not just research prototypes)
• Strong classical ML: XGBoost, LightGBM, CatBoost, time-series forecasting, clustering, SHAP-based interpretability
• Feature engineering depth: encoding strategies, temporal/lag features, interaction terms, leakage-free CV design
• Practical GenAI engineering: production RAG pipelines, chunking strategy, embedding model selection, retrieval quality metrics, reranking design
• Vector database fluency: Qdrant, Weaviate, ChromaDB, or pgvector — index types (HNSW, IVF), metadata filtering, hybrid search
• LLM orchestration: LangChain or LlamaIndex at production scale
• Prompt engineering as a discipline: version-controlled, systematically evaluated (chain-of-thought, few-shot, XML-tagged prompts)
• MLOps hygiene: MLflow/W&B, Docker, AWS basics (SageMaker, S3, Lambda), production monitoring
• HuggingFace ecosystem: transformers, datasets, sentence-transformers — able to benchmark and select embedding models

Preferred

• Hands-on MCP (Model Context Protocol) experience
• Fine-tuning experience: LoRA, QLoRA, PEFT on open-source LLMs
• LLM evaluation frameworks: RAGAS, LLM-as-judge, custom eval harnesses
• Experience with agentic workflows, tool schemas, multi-agent state machines (LangGraph or equivalent)
• Exposure to supply chain, procurement, logistics, or industrial B2B data/pricing/forecasting
• Familiarity with Ollama or vLLM for local inference benchmarking

Good to Have

• M.S./Ph.D. in CS, Statistics, Mathematics, or related field (or equivalent shipped-product depth)
• Multimodal AI experience — parsing PDFs/images (datasheets, BOMs), structured + unstructured extraction
• Open-source contributions to AI/ML libraries or published evaluations
• Familiarity with data governance, lineage tooling, or responsible AI frameworks (implemented, not just read)
• Experience with model quantization (GGUF, AWQ, GPTQ) or edge inference