Senior Data Scientist

Akaike Technologies Pvt. Ltd.; Bengaluru; On-site; INR 3,000,000 - 6,000,000; Full time

Job summary

Akaike Technologies seeks a Senior Data Scientist to architect agentic AI systems and deliver high-accuracy decision support for global clients in Bengaluru, India. The role blends Generative AI with traditional ML across large data, with room to explore problems as priorities shift.

Responsibilities include designing multi-agent systems, grounding, evaluation, and human-in-the-loop workflows, plus building production-grade pipelines on Databricks, AWS, and open-source models.

Qualifications

• Experience shipping agentic systems or complex RAG pipelines to production.
• Strong grounding, traceability, and hallucination control.
• Proven ability to design human-in-the-loop workflows for clients.

Responsibilities

• Design and ship production multi-agent systems with reliable outputs.
• Engineer grounding and source traceability for production use.
• Build evaluation metrics for generation quality and reliability.
• Design human-in-the-loop workflows scalable across clients.
• Build production RAG and PEFT/LoRA on open-source models.

Skills

GenAI frameworks

LangChain

LlamaIndex

LangGraph

DSPy

PyTorch

TensorFlow

PySpark

SQL

Python

Communication

Tools

Databricks

AWS Bedrock

AWS Lambda

AWS Step Functions

Job description

About the Role

Akaike Technologies builds agentic AI and machine-learning systems that power decision-making for global enterprises. We want a Senior Data Scientist who understands what makes agentic systems reliable and can architect complex, high-accuracy, human-in-the-loop AI for demanding clients. Roughly 70% Generative AI and agents, 30% classical ML and large-scale data, with room to move across problems as priorities shift.

Key Responsibilities

Generative AI & Agentic Systems (primary focus)

• Agentic Systems: Design and ship production multi-agent systems (LangGraph/CrewAI; ReACT, Agent-Critique) that generate reliable, structured, high-accuracy outputs.
• Grounding & Reliability: Engineer grounding, source traceability, and hallucination control so outputs hold up in production and regulated settings.
• Evaluation: Build quantitative evaluation for generation quality — faithfulness, hallucination, consistency, and LLM-as-a-judge frameworks.
• Human-in-the-Loop: Design configurable, human-in-the-loop workflows (review gates, co-pilot editing) that scale across clients and use cases.
• Retrieval & Tuning: Build production RAG and agentic retrieval; apply PEFT/LoRA on open-source models (Llama 3, Mistral) where it beats prompting.

Classical ML & Data at Scale (secondary focus)

• Custom Modeling: Build bespoke models for targeting, budget optimization, and churn, including PU and single-class learning on sparse, noisy data.
• Applied Deep Learning: Apply 1D/2D CNNs, LSTMs, and embeddings to non-text data (time-series, behavioral logs) where they add real signal.
• Big Data: Write optimized PySpark/SparkSQL on Databricks over billions of rows, with feature stores consistent across training and inference.
• System Design: Architect end-to-end systems with deliberate latency/cost/accuracy trade-offs and modular, reusable components.
• Measurement: Bring statistical rigor — offline and human eval, A/B where it fits, drift detection, and automated retraining.
• Deployment: Ship scalable pipelines on AWS (Bedrock, Lambda, Step Functions) and FastAPI.

Leadership & Stakeholders

• Mentorship & Clients: Mentor juniors, run rigorous code reviews, and serve as the technical point of contact for clients — explaining model limitations and risk without overselling.

Must-Have Skills

• Advanced GenAI/agent frameworks (LangChain, LlamaIndex, LangGraph, DSPy), with at least one agentic system or complex RAG pipeline shipped to production.
• PyTorch or TensorFlow; solid grasp of attention, encoder-decoder architectures, and embeddings.
• Expert PySpark and SQL; comfortable debugging and optimizing Spark on Databricks.
• Python (OOP, typing, rigorous standards) and a strong experimental / statistical mindset.
• Clear communicator who can explain complex AI — and its risks — to business leaders.

Nice to Have

• Pharma, life-sciences, or other regulated-domain background.
• Text-to-SQL, GraphRAG (Neo4j), or serving open-source models (vLLM/TGI).
• Custom loss functions / attention modifications, or open-source and publication contributions.
• Competitive ESOP grants.
• Working with Fortune 500 companies and world-class teams.
• Publishing papers and attending conferences.
• Networking events, conferences, and seminars.
• Visibility across all functions at Akaike — sales, pre-sales, lead generation, marketing, and hiring.

Required Skills

Data Science Machine Learning Deployment generative AI Agentic AI