Pritam Pandit — AI Researcher
I design end-to-end AI systems—retrieval, ranking, and LLM pipelines—that are observable, explainable, and cost-aware. Recent: productionized hybrid retrieval with multi-vector ranking, RAG, and LLM explainability; led a multi-agent POC; drove $100K NLP savings. Background includes Amazon (Alexa), Nielsen (media analytics), Ecolab (advanced analytics). IC-focused. NightOwl Consulting (AI Engineer): Built multi-vector ranking using Gemini embeddings (3072-d) with weighted fusion; hybrid retrieval in Qdrant (dense similarity + metadata filtering) and skill-overlap reranking; RAG with GPT-4 batch classification via structured prompts. Implemented LLM orchestration (JSONL, async polling, exponential backoff, auto-retry), NL query parsing with Gemini/OpenAI fallback, and LLM-powered explainability (rationale + evidence). Deployed on Render with FastAPI/Streamlit, GitHub Actions CI/CD, cron scheduling, and environment management; focus on latency, reliability, and observability. Vibe Engine AI (AI/ML Engineer): IC lead for multi-agent system (Selenium crawling → Pinecone embeddings → competitor SOV analytics) across closed LLMs; shipped agent-generated content and tracking that lifted AI visibility ~20%. Stack: Python, Pinecone, Selenium, prompt engineering, evaluation, dashboards. Amazon (ML Data Associate): Improved Alexa intent diversity by 10% and model output diversity by 40% via five novel data augmentation strategies across 50+ conversations; supported NLP/intent classification quality. Nielsen (Data Scientist): Increased pipeline efficiency 40% using AWS Glue + SQL transformations; delivered 10+ QuickSight dashboards over multi-TB datasets; reduced data quality issues 15% with NLP preprocessing/deduplication, saving $100K (cost optimization and data reliability). Focus on ETL/ELT, data modeling, and stakeholder reporting. Ecolab (Sr. Technical Engineer, Advanced Analytics): Built predictive analytics POC supporting a $16M initiative; deployed Power BI KPI scorecards that improved operational efficiency and compliance 30%; time-series risk modeling cut troubleshooting time 20%. Stack: Azure, SQL, Power BI, Python. Keywords and signals: AI Engineer, applied AI, LLMs (GPT-4, Gemini), RAG, hybrid retrieval, vector search (Qdrant, Pinecone), embeddings, ranking, prompt engineering, LLM orchestration, batch pipelines, observability, explainability, evaluation, CI/CD (GitHub Actions), FastAPI, Streamlit, AWS, Azure, SQL, ETL/ELT, data modeling, A/B testing, experimentation, cost optimization, stakeholder communication, production systems.
Stackforce AI infers this person is a SaaS-focused AI Engineer with expertise in data-driven solutions and predictive analytics.
Location: Manchester, Connecticut, United States
Experience: 9 yrs
Skills
- Ai Engineering
- Mlops
- Ai Content Generation
- Machine Learning
- Data Engineering
- Data Science
- Predictive Analytics
Career Highlights
- Designed end-to-end AI systems for retrieval and ranking.
- Led multi-agent systems improving AI visibility by 20%.
- Achieved $100K savings through data quality improvements.
Work Experience
NightOwl Consulting
AI/ML Engineer (11 mos)
Vibe Engine AI
AI/ML Engineer (5 mos)
Amazon
ML Data Associate (2 mos)
HeronAI
ML Engineer (2 mos)
Questrom School of Business, Boston University
AI Research Assistant (10 mos)
Nielsen
Data Scientist (1 yr)
Ecolab
Sr. Data Scientist (Advanced Analytics) (5 yrs 10 mos)
Education
Master in Business Analytics at Questrom School of Business, Boston University
Bachelor of Engineering (B.E.) at North Maharashtra University
