Brekfuz — Founding Engineer

Feb 2026 — present (2026—) · Full-time · Remote

Build Brekfuz as its first and only engineer: a knowledge layer that answers a question in the channel it was asked in, from the company's own Slack, Gmail and Calendar.

Capabilities

  • Permission-aware retrieval
  • Agent infrastructure

Outcomes

  • Built the answering layer, where source permissions are resolved before generation rather than after it, so one answer splits by clearance: the part the channel may see is posted in the channel with its sources attached, and the restricted remainder reaches the asker alone as a direct message. — as of 2026-09-06
  • Built the path for questions nobody has written down. The question routes to the person who made the call, their reply is captured, and the next person to ask gets it from memory instead of interrupting anyone. — as of 2026-09-06
  • Held the tenant boundary in the database rather than in application code: row-level security is forced on the tables that carry customer content, and the services verify it against the live catalogue at boot and refuse to start if it is missing. — as of 2026-09-06

Evidence: Brekfuz · Breken

Hey Amara — Founding AI Engineer

Aug 2025 — Jul 2026 (2025–2026) · Full-time · Remote

Build the backend and AI infrastructure of a multi-tenant agent platform for recruitment agencies, as its founding engineering hire.

Capabilities

  • Queue and messaging infrastructure
  • Retrieval and generation pipelines

Outcomes

  • Moved the platform's message-queuing layer off Redis and onto RabbitMQ, reworking the paths through it — circuit breaker, API gateway, orchestrator and service launcher — so agent tasks, tool results and streaming callbacks travelled over topic exchanges. — as of 2026-07-01
  • Built the email generation subsystem on a multi-source retrieval pipeline over vector similarity search, with a composition service, a generation orchestrator, and fetchers for recruiter memory and prior communications. — as of 2026-07-01
  • Built the auto-compaction that holds a long conversation inside its token budget by summarising earlier turns into a validated schema rather than a free-text prompt, and implemented the authentication middleware and the integration test suite around it. — as of 2026-07-01

Evidence: Hey Amara

UltraSafe AI — Applied AI Engineer

Aug 2025 — Oct 2025 (2025) · Internship · Dubai, United Arab Emirates · Remote

Build a medical retrieval system for regulated healthcare documents, from PDF extraction through to a deployed service.

Capabilities

  • Retrieval architecture
  • Production deployment

Outcomes

  • Built a seven-layer PDF extraction pipeline that runs six engines in parallel and reconciles them by vote, with coherence scoring, hierarchical chunking and OCR enhancement over clinical documents. — as of 2025-10-01
  • Built a multi-vector embedding framework fusing dense medical embeddings, SPLADE sparse representations and ColBERT token-level interaction, combined by weighted ensemble scoring over an HNSW approximate-nearest-neighbour index. — as of 2025-10-01
  • Built a four-stage retrieval pipeline — recall, cross-encoder reranking, knowledge-graph expansion with PageRank importance scoring, then redundancy filtering — and deployed it as Kubernetes services with Redis caching, a managed vector store and OpenTelemetry instrumentation. — as of 2025-10-01

Evidence: Medical RAG for regulated documents

Munshot — Founding AI Engineer

Mar 2025 — Aug 2025 (2025) · Internship · Singapore · Remote

Build an AI agent for institutional financial research, from its tool layer to the API that served it.

Capabilities

  • Tool server architecture
  • Agent reasoning

Outcomes

  • Built an MCP-compliant tool server putting more than ten tools behind one interface — a vector store, a general model, a search API and five proprietary data sources — for multi-source financial synthesis. — as of 2025-08-01
  • Built the agent's reasoning engine on LangGraph: a fifteen-node decision graph over more than twenty state keys that sequences tool calls, carries memory between them, and recovers from failure through retry logic and fallback parsing. — as of 2025-08-01
  • Built the FastAPI backend on non-blocking asyncio with JSON streaming for live agent progress and JWT-secured routes, and held the data contract at every API boundary and state transition with Pydantic models and their custom validators. — as of 2025-08-01

Evidence: A research agent for institutional finance

RealAIzation — AI Engineer

Jan 2025 — Mar 2025 (2025) · Internship · Texas, United States · Remote

Build the AI layer of a talent platform that ran the whole hiring loop, from writing the job description to reaching the candidate.

Capabilities

  • Agent orchestration
  • ATS integration

Outcomes

  • Built the orchestrator with adaptive strategy selection and reflection loops running over the multi-step hiring workflows. — as of 2025-03-01
  • Built a sourcing agent that queried five or more candidate data sources, with a scripting layer for custom scraping against niche talent pools. — as of 2025-03-01
  • Built the scoring engine: an auditable rationale attached to every assessment, bias guardrails, and semantic memory over prior hiring data. Integrations across three applicant tracking systems synced jobs both ways over webhooks. — as of 2025-03-01

Evidence: An AI platform that ran the whole hiring loop

HappyAspirants — Product Manager

Aug 2024 — Oct 2024 (2024) · Internship · Jaipur, India · Remote

Define and ship platform features for an interview-preparation product.

Capabilities

  • Feature specification
  • Competitive research

Outcomes

  • Led delivery of a live mentorship portal, one of two features taken from spec to release. — as of 2024-10-01
  • Led delivery of an integrated coding-challenge module aimed at the interview-preparation problems users actually reported. — as of 2024-10-01
  • Ran the competitive analysis and user research that defined the specs and a prioritised feature roadmap. — as of 2024-10-01

Springoo — Brand Manager

Mar 2024 — Aug 2024 (2024) · Internship · Jaipur, India · Remote

Build the brand for the company's main product and take it to market.

Capabilities

  • Brand development
  • Go-to-market creative

Outcomes

  • Led brand development for the company's main product, starting from nothing. — as of 2024-08-01
  • Coordinated designers, developers and editors so the work held together as one brand. — as of 2024-08-01
  • Owned the go-to-market creative: product packaging, visual identity, and the digital marketing and advertising campaigns. — as of 2024-08-01

The E-Cell, LNMIIT — Head of Sponsorship

Nov 2023 — Jun 2024 (2023–2024) · Jaipur, India · On-site

Raise the sponsorship that paid for the entrepreneurship cell's year.

Capabilities

  • Sponsorship strategy
  • Brand partnerships

Outcomes

  • Set the cell's sponsorship strategy and closed the partnerships against it. — as of 2024-06-01
  • Signed more than twenty brands across the cell's flagship events. — as of 2024-06-01
  • The sponsorship funded the cell's annual operations and its startup-summit programming. — as of 2024-06-01