Anand Altekar

I build systems that make operational intelligence tractable — before most organizations knew they needed it, and before most tooling existed to support it.

anand@decisionos.devdecisionos.dev →

Education

MS · Courant Institute, New York University

BSc Statistics · Fergusson College, Pune

Experience

Citi

INTERNSHIP

Reinforcement Learning & Deep Learning

Built a reinforcement learning trading system — custom Gym-style environments, reward design, backtesting loops, experiment tracking.

Policygenius

INTERNSHIP

Deep Learning at Scale

Pre-LLM, pre-ChatGPT: deep learning system to generate SEO content at scale from structured and unstructured inputs. End-to-end data pipelines, automated research, generation and quality assurance loops. The same architecture LLM applications now claim to have invented.

ControllerView · Axiom (now Nasdaq / Adenza)

Regulatory Reporting

Implemented regulatory reporting platforms for global banks — FFIEC 031, FR Y-9C, and others. Years inside the machinery made the path forward obvious: remove the manual abstractions, make the business logic explicit, automate reconciliation and audit steps.

Pushed the product in that direction. The organization wouldn't commit. I left.

Independent · Last 2 Years

LLM Systems & Reliability Infrastructure

Off-the-shelf agent and RAG stacks hit their limits fast — reliability gaps, no real debuggability, broken state management, no eval framework. Built my own, early, before most libraries offered robust patterns.

  • ·RAG with provenance tracking and knowledge graphs
  • ·Tool-calling agents: planning, routing, state, retries, fallbacks
  • ·Tool layer: SQL/DB, REST APIs, documents, web search, schema grounding
  • ·Guardrails and ops: permissions, audit trails, policy checks, deterministic checkpoints, monitoring, cost/latency instrumentation
  • ·Evals: test sets, regression harness, automated graders; Docker-first deployment

Decision OS

ACTIVE

Everything above led here. Decision OS is the system I kept wishing existed — built engagement by engagement, shaped by real enterprise complexity.

Schedule a Call →VIEW DECISION OS →

Currently taking 2–3 engagements per quarter.